https://orcid.org/0000-0001-6123-946X
Figures
Abstract
Understanding rainfall variability is important to establish crop calendar related agronomic decisions. To this end, we defined start and end of seasons, analyzed dry spell and evaluated conditional risks of alternative planting dates using a thirty years daily rainfall data across southern central rift valley of Ethiopia. Results showed that the probability of annual rainfall being greater than 1000 mm was 97, 24, 94, and 61%, in Dilla, Bilate, Shamana, and Hawassa clusters, respectively. The variability of annual total rainfall in the lowland areas of Dilla and Bilate was above 25%, whereas for Shamana and Hawassa was below 20%. Variability of seasonal rainfall during FMAM was 33.7%, which is higher than ONDJ (27.1%) and JJAS (27.9%), which could lead to maize plants suffering moisture stress during FMAM season. The onset of rains had variability of 29.2, 19.5, 17.5 and 26.5%, and also LGP showed variability of 22.8, 22.1, 21.2 and 20.3% in Shamana, Bilate, Hawassa, and Dilla clusters, respectively. Moreover Shamana, Bilate, Hawassa and Dilla clusters are hit by agricultural drought in one out of 2.61, 2.3, 2.5 and 2.5 years, respectively. Model based analysis of conditional risk of farmers planting dates also showed a success rate of less than 10, 7, 40 and 63% for maize variety in Shamana, Bilate, Hawassa and Dilla clusters, respectively. However, the success rate of risk taker farmers’ is higher than anticipated by the model. The farmers who take risk were encouraged in Shamana cluster by local edaphic, physiographic, socioeconomic and climatic differences. Hence, there is a need to seek real time local agro-metrological advisory and follow the necessary tactical and strategic farming decisions. Moreover, there is also a need to incorporate local factors with modern climate models to obtain synchronized calendar estimates.
Citation: Markos D, Worku W, Mamo G (2023) Spatio-temporal variability and rainfall trend affects seasonal calendar of maize production in southern central Rift Valley of Ethiopia. PLOS Clim 2(6): e0000218. https://doi.org/10.1371/journal.pclm.0000218
Editor: Erin Coughlan de Perez, Tufts University / Red Cross Red Crescent Climate Centre, UNITED STATES
Received: March 1, 2023; Accepted: May 16, 2023; Published: June 29, 2023
Copyright: © 2023 Markos et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper and its Supporting Information files.
Funding: The authors received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Abbreviations: CHIRPS, Climate Hazards Group Infra-Red Precipitation Stations; CV, Coefficient of variability; DOY, Days of the year; ENSO, El Niño–Southern Oscillation; EOS, Ending of the season; EMA, Ethiopian Mapping Authority; FMAM, February, March, April, May; HARC, Hawassa Agricultural Research Center; IOD, Indian Ocean Dipole; ITCZ, Inter tropical convergence zones; JJAS, June, July, August, September; EMA, Ethiopian Mapping Agency; EMI, Ethiopian Meteorological Institute; NMSA, National Meteorological Services Agency; OND, October, November and December; ONDJ, October, November, December, January; QBO, Quasi-Biennial Oscillation; SOS, Start of belg season; SARI, Southern Agricultural Research Institute; SPI, Standardized precipitation index; XLSTAT, Excel data analysis add-on
1. Introduction
Understanding the year-to—to-year variation of rainfall variability and pattern is important not only for rain-fed crop production but also for designing wise use of water resources in hydrology, ecology, irrigation, and agriculture [1–3]. This is particularly important in Ethiopian agriculture, which is being strongly affected by the risks associated with rainfall variability. More expressively, the unreliable rainfall distribution combined with the high evaporative demand of the tropical climate results in a high risk of water deficit at any stage of crop growth [4].
Given that Ethiopia is a distinctive and most rugged country in Africa, rainfall in Ethiopia is complex by its very nature and substantially modified by local factors in which the topography is powerful [5,6] where mountain ranges create barriers that alter wind and rainfall patterns. The prevailing winds are affected by mountains and hills form micro-scale pressure systems, which create moisture flow as a rain shadow on their leeward (protected) sides, where the air contains very little moisture [6], resulting in variations in rainfall direction, seasonality, rainfall onset, amount, cessation dates, and length of growing season [7,8]. In addition to local factors, the moist air flowing from the Indian Ocean and Gulf of Aden causes a northward advance of the Inter tropical Convergence Zone (ITCZ) towards low pressure areas over Arabian and Southern Sudan producing Belg (February–May) rains, that is restricted to east, southern and southeast parts of Ethiopia due to orographic barriers [9]. Whereas during June–September, the moisture-laden westerly winds flowing from the Atlantic and Indian Oceans towards the extreme northern Ethiopia [10], produce rains over most parts of the country, except for the drier conditions over the southern and southeastern lowlands. During October-November-December-January (ONDJ), the country predominantly falls under the influence of dry and cool northeasterly winds originated from the Saharan and Siberian anticyclone, which forces the southward migration of ITCZ causing the rainfall over Ethiopia to retreat towards the south, thus providing small rains in the southern part of the country [11].
Regional and local weather influencing systems play key roles in producing rainfall of differing pattern in Ethiopia. Globally, the Pacific ocean and El Niño Southern Oscillation (ENSO), regionally Indian and Atlantic oceanic interaction process, Indian Ocean Dipole (IOD), other large scale atmospheric phenomena including the upper level winds such as Quasi-Biennial Oscillation (QBO) and high pressure systems in the Indian and Atlantic oceans were reported as causes of variability in rainfall pattern and distribution across locations [12]. The global and region weather systems are affected locally due to topographic variations and geographic locations rendering the country into four rainfall regimes of differing rainfall occurrence [5,13,14]. These rainfall regimes are shown in Fig 1 as 1) Regime A or bimodal type-1 (quazi double maxima), 2) Regime B or mono-modal (single maxima), 3) regime C or bimodal type-2 (double maxima) and 4) regime D with diffused rainfall pattern. Regime A is characterized by a quazi double maxima rainfall patterns, with one peak rainfall in April and another peak in August. Indeed, this regime exhibits semi-bimodal rainfall pattern. The wet period in regime B decreases northwards. Thus, regime B in turn is re-clustered into three parts designated as area b1 (February/March to October/November), area b2 (April/May to October/November), and area b3 (June/July to August/September) [5,14–16]. Regime C is dominated by a double maximal rainfall pattern. When the ITCZ is situated at its north most position, regime C remains dry (June to September), whereas when the ITCZ retreats from the northern towards the southern hemisphere, by virtue of the system’s nearness, Regime C experiences its second rainfall or a relatively short rainfall during October, November and December (OND). Regime C also experiences relatively long rains during February-March-April-May (FMAM).
Thus, the southern part of central Ethiopia, which is a maize production belt of the country, clustered into the southern part of regime A and the northern part of regime C. In this area, once the growing season begins, extended dry spells influence water availability [5], thus, becoming a challenge across the critical growth stages of maize crop. Dry spell risks result in weakening of crop roots and their nutrient absorption capacity, thus leading to soil conditions where soil water cannot afford to sustain crop growth and development at critical growth stages [17–20]. This also exposes the standing maize plants to an increased risk of stalk borers, aphids, and associated virus diseases, and therefore, reducing crop productivity [21]. Kingra et al. [22] developed evidences on the role of adjusting crop sowing dates and subsequently avoiding the coincidence of extended dry spells with sensitive crop growth stages such as emergence, flowering or grain filling. Hence, knowledge on extended dry spells is increasingly growing for designing mid-season management practices such as tie ridging, mulching, reducing plant population, defoliation and supplemental irrigation, which are useful to reduce its adverse effects on maize production in the southern region. This knowledge base is vital to develop alternate land use and soil water management practices to be bundled with other services, like agro-advisory services that help informed decision making to leverage existing or new water conserving technologies [23–26].
Scientific evidence related to the southern part of regime A and the northern part of Regime C is scanty. It is only simple descriptive statistics like mean and standard deviation, which does not provide a well-defined characteristic distribution of in-season extended dry spells and spatiotemporal variability that has been used [27–29]. Hence, the results are inconsistent; lacking site-specific information on the spatial and temporal variability, and seasonal and annual rainfall trends [4,6,24]. Therefore, this study was aimed (1) at exploring the spatio-temporal trends of onset, cessation, duration, and amount of rainfall, together with dryness, wetness, and dry spells that have been found necessary for understanding rainfall characteristics and (2) to make informed decision on maize seasonal calendar and associated management practices across study clusters.
2. Materials and methods
2.1 The study area
The study area lies within 6.38 to 7.72 (decimal degrees) latitude and 37.74 to 38.868 (decimal degrees) longitude (Fig 2), and comprises four homogeneous clusters (Table 1). The source of the base map shape file was Ethiopian Mapping Agency. The model results were exported into ArcGIS Sofware Version 10.2 (http://desktop.arcgis.com/en/arcmap) to generate the map in Fig 2.
The source of the basemap shape file was Ethiopian Mapping Agency. The model results were exported into ArcGIS Sofware Version 10.2 (http://desktop.arcgis.com/en/arcmap) to generate the map in Fig 2.
2.2 Data source
Historical daily rainfall records covering a period of thirty years (1991 to 2020) for each station located within and in close proximity to the study area were obtained from different sources; including the Ethiopian Meteorological Institute (EMI), Hawassa Branch office of EMI, Hawassa Agricultural Research Center (HARC). In areas where ground observation were lacking, Climate Hazards Group Infra-Red Precipitation Stations (CHIRPS) satellite rainfall data (https://data.chc.ucsb.edu/products/CHIRP S-2.0/) [12,30] were used. Subsequently, seasonal and annual time series data were computed as required for variability and trend examination (Table 1). Filling gap was made where ever missing data was below 10% using XLSTAT statistical software [31]. The examination of homogeneity and normality was done using the cumulative deviation method and no heterogeneity was detected.
2.3 Rainfall variability
2.3.1 Recurrence of rainfall.
The temporal variability and occurrence of rainfall were evaluated at selected weather stations based on the analyses of a set of indicators defining variation and extreme conditions following Stern et al. [32]. The probability of exceedance analysis were used in which rainfall amount data were ranked in descending order and given a serial rank number (r) ranging from 1 to n (number of observations). Then, we plotted the amount of annual rains in x-axis and the probability of exceedance of the total rainfall on y-axis. The probability of exceedance (Pi), expressed as a fraction (on a scale ranging from zero to one) was computed using Weibull Eq 1 [33,34].
Where n represents the number of days of rainfall and i indicates the rank.
2.3.2 Rainfall onset and cessation.
Rainfall onset and cessation date variability indicators were determined using R-Instat v 0.7.2 software [32,35].
- Criteria for the start of belg season (SOS)
The first occasion with a rainfall total of 20 mm or more and a length of consecutive 3 days or more; no dry spell duration of 10 days of consecutive or more in the next 30 days after February 1st (Clusters I and IV), April 1st (cluster II) and March 1st (cluster III) [26,35].
- Cessation of the rainy season (EOS)
The earliest possible day of May 1(for Cluster I) or September 1 (for II, III, and IV) when the capacity of soil to persist precipitation with the water balance equals zero [35].
2.3.3 Length of rainy season.
Length of the rainy season was obtained as the duration in days between onset and cessation dates. Rainfall characteristics were captured across the whole growing period of the crop procedures contextualized by Kassie et al., [26] and Mamo [36]. The maize crop starts growth in one season and harvested before the end of another season, with exact operational calendar varying greatly across clusters. The 80, 50, and 20% probabilities of exceedance of the start and end of the growing season are used as indicators of early, normal, and late onset and cessation of rains [37]. The same information was use to determine length of growth period cross clusters.
2.3.4 Dry spell analysis.
Dry spell analysis was performed using the 1st order Markov chain probability model to fit sequences of dry days in records of varying lengths and for several climatically different areas [37,38]. The first order (two-state) of the model has been applied to determine the initial and conditional probabilities of dry spell weeks. The denotations used for dry spell lengths were: (sp7) 7 days, (sp10) 10 days, (sp15) 15 days, and (sp25) 25 days varying from place to place. The average dry spell length was determined based on a 1 mm threshold [39]. Hence, the occurrence of dry spells within a season was identified using a threshold of less than 1mm per day within a season [35]. For the analysis of the probability of dry spells, the following parameters were estimated as described in [40] for the initial probabilities (Eqs 2 and 3):
Eq 2
Eq 3
For conditional probabilities (Eqs 4 to 7)
Eq 4
Eq 5
Eq 6
Eq 7
Where Pd is the probability of a week being dry; Pw is the probability of a week being wet; Fd is the number of dry weeks, Fw is the number of wet weeks, N is the number of years of data (30 years), Pdd is the probability of a dry weeks preceded by a dry week, Pww is the probability of a wet week preceded by a wet week, Pwd is probability of a wet week preceded by a dry week, Pdw is probability of a dry week preceded by a wet week, Fww is number of wet weeks preceded by a wet week, Fdd the number of dry weeks preceded by a dry week.
2.3.5 Wetness and dryness of the area.
The wetness and dryness condition of each cluster was studied using the standardized precipitation index (SPI), which is the most widely used tool of drought index to detect and characterize agricultural, hydrological, and meteorological drought [41]. Hence, the classification of dryness and wetness in accordance with SPI was made for a period of 1–3 months to study dryness and wetness related to agriculture. The SPI is equivalent to the z-score often used in statistics (Eq 8):
Eq 8
where Xij is the seasonal precipitation at the ith rain gauge station and jth observation, where Xim is the long-term seasonal mean, and where σ is its standard deviation. A positive value of SPI indicates that precipitation is above average and a negative value shows below average. Values of (−1 to +1) express a normal precipitation regime and values out of this range represent relevant deviations from the normal rainfall amount. Consequently, the values of SPI were noted as extremely wet (> 2.0), very wet (1.5 to 1.99), and moderately wet (1.0–1.49). Values are moderately dry if belong to (−1.5 < SPI ≤ −1), severely dry (−2 < SPI ≤ −1.5), and extremely dry (SPI ≤ −2.0). A drought event starts when SPI value reaches -1.0 and ends when SPI becomes positive or closes to positive again [41].
2.4 Trends of rainfall
The seasonal rainfall, its characteristics, and SPI time series data were evaluated using Mann–Kendall’s test statistics at a 5% significance level to detect the prevalence of monotonous trends, and determined by making continuous autocorrelation using the Yue and Wang method to detect serial correlation and seasonality effects after sorting the time series data in ascending order prior to analysis [42,43]. The null hypothesis (H0) and alternative hypothesis (H1) were stated as
Ho: The rainfall values were randomly ordered in time i.e independent (no temporal correlation)
H1: The rainfall values had non- reversing increasing or decreasing trend (single direction)
The test statistic (Kendall’s S) is defined as in Eq 9:
Eq 9
where sign (θ) is the sign function that equals -1 when θ < 0, 0 when θ = 0, and 1 when θ > 0, n is the number of data in X, j,…, k…, s…. S is asymptotically normally distributed under the null hypothesis when n>10. If there is no trend, then the null hypothesis is true and conversely if there is any trend, the null hypothesis is false. Kendall’s S would then be expected to be close to 0 whenever there was no trend. If S is significantly different from zero, the data indicate that a trend in Y has occurred, indicating a positive or negative single direction trend (Eq 10).
Eq 10
where n is the length of the sample, xj and xk are from k = 1, 2, …, n-1 and j = k+1, …, n. If n is bigger than 8, the statistic S approximates to a normal distribution. The mean of S is 0 and the variance of S can be acquired as follows (Eq 11):
Eq 11
Kendall’s Tau (τ)—is a nonparametric (independent and less sensitive to outliers) correlation coefficient that measures the strength of the monotonic relationship between rainfall and time in this study. Tau is a rank-based procedure and is therefore resistant to the effect of a small number of unusual values. Because τ depends only on the rank of the data and not the values themselves, there are adjustments for missing or censored data (essentially treated as ties)–tests work with a “limited number of” such data as shown in Eq 12:
Eq 12
where τ = a nonparametric rank order (pair-wise comparison) correlation statistic between observations and time series data, S is the number of intersections, n is the number of observations [44].
τ = -1 indicates a perfect negative monotonous relation between 2 variables
τ = 0 indicates no monotonous relation at all and
τ = 1 indicates a perfect positive monotonous relation: a lower score on variable A is always associated with a lower score on variable B. For larger sample sizes, S is converted to Z, a statistic that is approximated by a normal distribution. The formula for Z is given below in Eq 13.
2.4.1 Sen’s slope estimation.
Sen’s nonparametric method gives a robust estimation of the time series trend and used to estimate the magnitude of trends in the rainfall time series data [45], whereas multiple estimates (N’) are made of the slope using Eq 14:
Eq 14
Where β is Sen’s slope estimate, x …., i…., j….. Sen’s slope estimate of β>0 indicates upward trend in a time series. Otherwise, the data series presents a downward trend during the period.
2.5 Analysis of conditional risk
Conditional risk of alternative dates of planting was analyzed by using R-instat software [32] with the following treatment combination for each cluster.
Shamana cluster—maturity period of variety: 120, 150 and 180 days; planting days (DOY):20, 40, 60, 80, 100 and 120 and water requirement (mm): 400, 500 and 600
Bilate cluster—maturity period of variety: 100, 110, 120, 130 and 140 days; planting days (DOY):20, 40, 60, 80, 100 and 120, and water requirement (mm): 300 and 400
Hawassa cluster—maturity period of variety: 120, 150 and 180 days; planting days (DOY):20, 60, and 100, and water requirement (mm): 400, 500 and 600
Dilla cluster—maturity period of variety: 120, 150 and 180 days; planting days (DOY):20, 60, and 100, and water requirement (mm): 400, 500 and 600
2.6 Crop-climate seasonal calendar
Seasonal crop-climate calendar was sketched to show temporal information of a particular cluster by keeping the space factor (clusters) constant and varying the time factor (months of the year) as variable. In a given cluster, the rainy season begins with SOS and ends with EOS (inclusive of short rains, long rains, short dry, and long dry period) which were obtained from R-Instat v 0.7.2 software [32,46]. The cropping season starts with emergence and ends with the physiological maturity of the crop. The seasonal calendar begins with agricultural operations like land preparation, planting, weeding, urea application etc. and ends with harvesting [47]. The rainy season, cropping season and seasonal calendar collectively formed the crop-climate calendar of an area. These three were triangulated against secondary and primary data obtained from office of agriculture and key informant farmers’, respectively across the study clusters.
2.7 Ethics statement
Experts and farmers were informed about the purpose of the study and asked for their verbal consent to participate prior to the calendar interviews. It was stated that the data collected would be treated confidentially, analyzed anonymously, and used only for research purposes. Their participation in the survey did not involve any risk to farmers.
3. Results and discussion
3.1 Rainfall total and its recurrence
For Dilla station (cluster IV), the 10th, 25th, 50th, and 90th percentile of annual rainfall was 1037, 1279, 1384, 1396, 1504, and 1769 mm, respectively (Fig 3). For Hawassa station (cluster III), the minimum, maximum, 1st quartile, median, mean, 3rd quartile rainfall was 769, 1288, 901, 1023, and 1102 mm, respectively. In Bilate station (cluster II), the 10th, 25th, 50th, and 90th percentile of annual rainfall was 682, 820, 894, and 1257 mm, respectively. For Shamana station (cluster I), the minimum, maximum, 1st quartile, median and 3rd quartile rainfall was 887, 1652, 1101, 1206, and 1272 mm, respectively.
Exceedance probability graph (left) and box plot (right) of rainfall for the study area.
The mean rainfall of thirty years (1991–2020) in representative stations of Bilate, Shamana, Hawassa, and Dilla clusters was 912, 1218, 1017, and 1396 mm, respectively. Thus, the highest mean annual rainfall was recorded in cluster IV at Dilla station and the lowest mean annual rainfall was recorded in cluster II at Bilate station (Fig 3). The rainfall in clusters I and III was intermediate in amount. However, only 1 year out of the thirty years received 1510 mm of rainfall in Shamana cluster and for fifteen years out of thirty, the Shamana area received 1208 mm of rainfall (Fig 3 exceedence graph). In Bilate cluster, only one year received 1228 mm of rains whereas fifteen years received 897 mm of rains out of thirteen years. The total difference between the highest annual rainfall (about 1228 mm) and the lowest (about 690 mm) is 538 mm. This value indicated an increased percentage of 228.3% between the two. The observed large deviation of seasonal rainfall was an indication of high variability and associated risk for maize production in these areas. Areas in Hawassa cluster received rainfall of 1288 mm in one out of thirty years (Fig 3 exceedance graph). However, over and above 877 mm of rainfall was recorded in Hawassa area every year. Areas in Hawassa cluster received at least 1018 mm of rainfall in one out of two years.
In Dilla cluster, only one out of thirty years experienced the maximum rainfall of 1769 mm, almost every year there was 943 mm of rainfall, and in one out of two years Dilla received 1395 mm of rains. However, the Dilla cluster showed the largest variability in seasonal rainfall with a range of 826 mm. Moreover, Bilate, Shamana, Hawassa, and Dilla clusters showed inter-quantile range of 159, 171, 201, and 225 mm, respectively (Fig 3. Box plot). Thus, there is temporal variability of rainfall across all clusters, even more so in those receiving higher rainfall, which would probably result in a high risk of growing maize under normal conditions. Hence, there is a need to seek for agro-metrological advisory and follow the necessary tactical and strategic farming decisions.
Thus, the probability of the total rainfall being greater than 1277 mm of rain (Fig 3. exceedance graph) is 74, 0, 26, and 4% in Dilla, Bilate, Shamana, and Hawassa areas, respectively. Moreover, the probability of the total rainfall being greater than 1000 mm was 97, 24, 94, and 61%, respectively. Thus, there was a greater than 50% probability of getting 894, 1208, 1016, and 1386 mm of rainfall every year in Bilate, Shamana, Hawassa, and Dilla clusters, respectively.
3.2 Onset, cessation and length of rainy period
Cluster I (Shamana cluster)—There was strong interannual variability in the mean onset and cessation dates of rainfall, subsequently leading to variability in the length of the rainy period over years. The rains start in the 3rd dekad of March at Boditti and shone areas of cluster I. The onset of Belg is highly variable, with standard deviations of 21.9 and 21.7 days in Shone and Boditti areas, respectively. The end of the rainy season occurred on the 2nd dekad of September and August with standard deviations of 33.4 and 37 days, respectively (Table 2). The relatively high standard deviation of the end of the season ranging from 33.4–37.0 days over southern- central Ethiopia is partly a reflection of the relatively short distinct dry period. Length of the rainy season ranged from 184 days in Shone to 156 days in Boditti area of the Shamana cluster. The coefficient of variability of SOS was 30.2 and 29.2 in Shone and Boditti areas, respectively, whereas the coefficient of variability for EOS was 14.4%. Variability’s of the onset of rainfall can be considered as the major problem in this cluster, and hence agro-met advisories will become essential. The LGP had CV of 23.5 and 22.8% in Shone and Boditti areas, respectively. This onset result showed a forward shift of two dekad within 16 years compared to reports of Mesay [48] who showed 29 February as a potential rainfall onset date.
Cluster II (Bilate cluster)—The rains start in the last dekad of March in both Halaba and Aje areas (Table 2). The onset of Belg is highly variable, with a standard deviation of 19.5 days in Halaba and Aje areas of Bilate cluster. The end of the rainy season occurred on the 3rd dekad of July in Halaba and Aje areas with standard deviation of 16.2 days. The relatively high standard deviation of the end of the season ranging from 21.1–18.9 days over southern- central Ethiopia is partly a reflection of the relatively short distinct dry period. Length of the rainy season ranged from 128 days in Halaba to 124 days in Aje area of the Bilate cluster (Table 2). Thus, there may not be enough time to grow hybrid varieties of maize that have maturity duration longer than four months in this cluster. The coefficient of variability of SOS was 19.5 in both Halaba and Aje areas, whereas the coefficient of variability for EOS was 10 and 9.1%, respectively. The LGP had CV of 22.1 and 21.2% in Halaba and Aje areas, respectively. The rift valley regions generally have short growing seasons, which are manifested by late onset and early cessation dates. The areas in cluster II had the relatively shortest length of rainy period that corresponds to 124 to 128 days. This result was in harmony with Wodajo et al. [49] who obtained the onset in 102 DOY, and endings of DOY 269 and 194, and the length of rainy days of 168 and 92 for Halaba and Bilate areas, respectively.
Cluster III (Hawassa cluster)—The rains start in the last dekad of March in Hawassa and Yirba areas of Hawassa cluster (Table 2). This onset result deviated with reports of Mesay [48] who showed 10 March as the onset in Hawassa area. The forward shift of onset could be attributed to climate change in the area. The variability of onset of Belg was shown with a standard deviation of 20.5 and 14.6 days in Hawassa and Yirba areas of Hawassa cluster, respectively. The end of the rainy season occurred in the 3rd dekad of August in Hawassa and Yirba areas with standard deviations of 29.5 and 35 days, respectively. Length of rainy season was about 140 days both in Hawassa and Yirba areas of the Hawassa cluster. The coefficient of variability of SOS was 23.7 and 17.5 in Hawassa and Yirba areas, respectively, whereas the coefficient of variability for EOS was 12.5 and 14.9%, respectively. The LGP had CV of 21.2 and 26.5% in Hawassa and Yirba areas, respectively.
Cluster IV (Dilla cluster)—The rains start in the 2rd dekad of March in Dilla and Aletawondo areas of cluster IV. The onset of Belg had a standard deviation of 20 days in Dilla and Aletawondo areas of Dilla cluster. The end of the rainy season occurred in the 3rd dekad of August and 1st dekad of September in Aletawondo and Dilla areas with standard deviations of 37.0 and 35.3 days, respectively (Table 2). Length of the rainy season ranged from 163 days in Dilla to 172 days in Aletwondo area of the Dilla cluster. The coefficient of variability of SOS was 26.5 and 28.3 in Aletawondo and Dilla areas, respectively, whereas the coefficient of variability for EOS was 15.6 and 15.0%, respectively. Variability’s of onset of rainfall can also be considered as the major problem in this cluster too, and hence agro-met advisories will become essential. The LGP had CV of 22.3 and 20.3% in Dilla and Aletwondo areas, respectively. According to Mesay [48], the SOS was 24-February whereas the EOS was 3rd September. Thus, after 16 years of Mesay’s study, the onset had shifted forwards by 3 dekad and the end dates were shortened by two dekad, by drastically shortening the rainy period in Dilla cluster. Earliest onsets occurred during the years of El Niño and usually the onsets were associated with SST and easterly winds in this area, which agrees with Hachigonta et al. [50] who identified the onset and cessation dates of the main summer rainy season in Zambia.
The LGP showed the high level of variability in all clusters. This is due to shifts in the onset and end of growing seasons, which is due to changes in temperature and rainfall patterns in the area. The higher variability in LGP is in agreement with the reports of Natai [51] who obtained similar results for three locations in Central Rift Valley. Using Reddy’s (1990) classification, the length of rainy season was moderately stable in Bodiiti, Dilla, Aletawondo, and Hawassa with standard deviation ranging between 30–40, whereas the length of rainy season was least stable in Shone and Yirba with standard deviations higher than 40. In fact, the CV of the length of growth period ranged between 20 to 30%, which confirmed the increasing difficulty in selecting varieties based on the maturity period across clusters [52]. The observed high variability in the length of rainy season calls for coping mechanisms through selection of medium to early maturing varieties, implementation of rainwater water harvesting scheme, and application of conservation tillage practices to ensure sustainability of maize production.
3.3 Trends of onset, cessation, and length of rainy period
The start of belg rainfall showed a significant (P<0.05) decreasing trend of 1.82 days/year in Boditi area of Shamana cluster (Table 2). This is a manifestation of the increasingly late onset of rains over the last thirty years in transitional to highland maize growing areas. The start of belg rains showed significantly increasing tends in Hawassa and Shone, but non-significant trends in other stations. The cessation of rains had no defined pattern across the stations considered. The duration of rainfall had a decreasing trend at Shone, Halaba, Aje, Hawassa, Yirba, and Dilla at the rate of 0.182, 0.333, 0.5, 0.92, 0.2 and 0.7 days/year, respectively. This showed a shortening of the growing period leading to a narrowing windows for crop production. The decreasing trend of rainfall duration was significant (P<0.05) at Hawassa station, which agreed with the reports of Thompson [53] who predicted a decline in the number of days suitable for growing crops in East Africa. In fact, predicted climate change represents an enormous challenge that will test farmers’ ability to adapt and improve their livelihoods and calls for practical strategies that farmers can use to build intrinsic resilience within maize production systems.
3.4 Seasonality of rainfall
Shamana cluster- In Bitena area, the first peak occurred in May with 174.1 mm of rain/month, whereas the second peak occurred in August with 150.1mm/month (Fig 4). In Shamana, the first peak occurred in May with 147.9 mm/month, whereas the second peak occurred in September with 121.3 mm/month. Below 50 mm/month was measured in both Shamana and Bitena areas in November and December. As these areas are located in transitional highland topography, the evaporative loss of incoming rains was comparatively lower due to subsequent cooler temperatures and higher humidity. This becomes an opportunity for maize cultivation.
Bilate cluster–The first peak occurred in May and April with 118.6 and 134.6 mm/month in Falka and Bilate areas, respectively (Fig 4). There was a tendency of the second peak to decrease over time. The areas in this cluster received below 100 mm/month between mid-July and mid-March. The amount of mean monthly rainfall received during the peak months was low and insufficient to support maize planting in this cluster mainly because such light falls do not penetrate deeply into the dry topsoil and are quickly evaporated in the succeeding dry spells. Hence, farmers wait until the last week of April and the first week of May for planting while preparing their land with available moisture until then. However, lately arriving rain also breaks in the latter growing season of maize causing big loss. Overall, assessment depicted that three to four month maize cultivars should be adopted in Bilate cluster.
Hawassa cluster–In Ropi area, the first peak occurred in May with a monthly total rainfall of 153.4 mm, whereas the second peak occurred in August with 142.9 mm/month (Fig 4). In Hawassa area, the first peak occurred in May with a monthly total rainfall of 135.8 mm, whereas the second peak occurred in August with 136.5 mm/month. In Hawassa area, the period between October and May was dry and risky to grow any crops without irrigation. Hence, farmers used the moisture obtained in March for land preparation and did planting of maize by the end of April without waiting for peak rainfall to occur. The main rain received in JJAS was associated with a period of maximum weather activity in the region during these months as a result of the movement of Intertropical Convergence zone (ITCZ) oscillating from West Africa towards India and leading to the development of a tropical easterly jets [54–56]. Thus, Shamana, Bilate and Hawassa clusters belonged to the rainfall regime A of Ethiopia with the first peak in April/May and the second peak in August/September period of the year.
Dilla cluster–The first peak occurred in May in Dilla and Aletawondo areas with 239.8 and 194.1 mm/month, respectively. The second peak occurred in October with 207.7 and 195.9 mm/month in Dilla and Aletawondo, respectively. In Dilla area, the risky periods occur twice between December and March, but also between June to August. Although the non-seasonal rainfall in the short and long dry periods brings temporary relief from drought, maize plants suffer from moisture stress that usually causes direct economic loss to farmers. Thus, Dilla cluster belonged to regime C or bimodal type-2 of Ethiopian rainfall as the first and second peaks of rainfall coincide with the teleconnections of rainfall for southern Ethiopia (Fig 4).
The source of the base map shape file was Ethiopian Mapping Agency. The model results were exported into ArcGIS Sofware Version 10.2 (http://desktop.arcgis.com/en/arcmap) to generate the map of four clusters in Fig 4.
3.5 Spatial rainfall variability
The associated coefficient of variation (CVs) for annual rainfall was 13. 1% in Abaro and 16.1% in Shamana area of cluster I. In cluster II, areas of Bilate and Abaya, CV of 26.4% and 27.5%, respectively, was obtained (Table 3). The CV of annual rainfall was 16.1% and 21.5% in cluster III areas of Hawassa and Wolaita sodo, respectively. Cluster IV areas of Aletawondo and Dilla had CV of 12.9% and 23.4%, respectively. Using the indices set for rainfall variability (Hessebo et al., 2019), all locations with CV measures higher than 25 were also reported to show higher temporal inter-annual variability of rainfall and known to be vulnerable to drought. The CV for annual total rainfall in the lowland areas of Dilla, Bilate and Halaba was above 25%, whereas for the highland areas of Shamana and Abaro was below 17%. Thus, the variability of rainfall was highest in moisture stressed lowland areas whereas lowest in the high altitude areas. The rainfall variability in FMAM was above 33.7%; in JJAS was 27.9% and in ONDJ was 27.1%. Thus, the highest seasonal rainfall variability was observed in FMAM season (Table 3). In Dilla cluster, the contribution of FMAM rains (35.3%) to annual rains was higher than the other two seasons. In other clusters, the contribution of JJAS rains to annual rainfall was higher than either ONDJ or FMAM rains. Although JJAS season contributed about 41.3% of the annual rainfall, there was 30.9% and 27.8% contribution from ONDJ and FMAM, respectively, in the study area (Table 3).
Overall, JJAS season had a major contribution to annual rainfall in all clusters compared to other seasons. This result is lower than the finding of Krauer [57] who identified JJAS season to contribute over 50% of the annual rainfall in the country. The results depicted that the contribution of JJAS rains to annual rainfall in Southern Central Rift Valley were lower than that of national and regional average, which were 66 and 55%, respectively [58,59]. However, only transitional highland areas of cluster I utilize that moisture during ONDJ period for planting maize. In other clusters, ONDJ rains had little agricultural value. The variability and contribution of rainfall computed for the whole Southern Central Rift Valley showed the highest variability (33.7%) and contribution (41.3%) in FMAM and JJAS, respectively. Such aggregation did not recognize extreme variations in each cluster and subsequently mislead management actions. Moreover, intrannual (seasonal) rainfall showed higher CV compared to interannual rainfall (Table 3) depicting higher variability across seasons compared to that of across years.
3.6 Monthly and annual trends of rainfall
The trend of rainfall showed increasing, decreasing, or no trend patterns across clusters. In Dilla area, strongly decreasing trends in annual rainfall were observed (τ = -0.251, S = -109) (Table 4). The rate of decrease of annual rainfall in Dilla cluster was 4.21mm/decade. Conversely, an increasing trends of annual rainfall were observed in Bilate (τ = 0.154), in Ropi (τ = 0.148), and Hawassa (τ = 154). This result is in harmony with the studies of Muluneh [24] who obtained the increasing trend of annual rainfall in the semi-arid areas at the floor of Central Rift Valley. There was increasing trend of annual rainfall in Aletawondo, Abaro, Bilate, Ropi, and Hawassa at the rate of 9.4, 6.1, 5.7, 3.3 and 0.24 mm/year, respectively. These findings were in line with Befikadu [60] who reported a significant upward trend in the annual rainfall in the midland agro-ecology of Southern Ethiopia and to those of Wodajo et al. (2016) who obtained increasing rains at the rate of 4.76 mm/year in Bilate station. This finding is dissimilar to the results of Befikadu [60] who found a non-significant trend for annual rainfall across years in lowland and mid-altitude transitional areas. Only the station in Halaba resulted in non-significant (P<0.05) yearly trend of rainfall. This result agreed with Hessebo [61] who found similar trends for rainfall for Halaba and Shone areas. The declining trend of rainfall in Dilla cluster was in agreement with Cheung [62] who reported a significant declining trend of annual rainfall in the nearby Yirgachefe area. This result is also in support of Viste [63] and Makin [64] who indicated the decrease of rainfall from 1971 onwards. There was a decreasing trend of monthly rainfall in Dilla cluster for ten out of twelve months (except May and October). This confirmed the reports that rainfall has been declining constantly since the 1980s, with the last few years being particularly dry in FMAM and ONDJ rains receiving areas of the southern, south-central, and south-eastern parts of the country [65].
For the months of January, February, March, and May, there was a similar decreasing trend of rainfall in Shamana and Hawassa clusters (Table 6). This agreed with reports of Ayele [66] who revealed a decreasing trend of rainfall in six out of twelve months (January, February, July, August, September, and October) in Cluster III area of Hawassa. Months of January, July, September, and December showed a decreasing monthly trend of rainfall even in Bilate areas, which was indicated in this and many other reports with increasing annual and seasonal rainfall [16,60,61]). This showed that seasonal or annual increases were attributed to increases of very few peak periods only. However, there was not any trend of monthly rainfall in Shamana, Bilate and Hawassa clusters for months of February, March, or May. The lack of vivid trend during the periods of planting, urea application and weeding of maize poses a great deal of challenges in making farming decisions. Thus maize plants suffered moisture stress in Shamana during January to March, Bilate in July, Hawassa in June and Dilla in March, April and June. Elias and Yaya [17] had also shown decreasing trends in autumn and spring in cluster I Boditti area. Maize plants used to suffer decreasing rains over the last thirty years from January to March at Shamana, in July at Bilate, in June at Hawassa, and in March, April and June at Dilla areas.
3.7 Dry spell probability
Shamana cluster—The probability of dry spell length of greater than 25 days was below 40% even in January (Fig 5). The probability of dry spell length greater than 7, 10, and 15 days was more than 50% prior to the 18th of March, 7th of March, and 19th of February, respectively (Fig 5). The probability of dry spell length greater than 7 days was below 10% between 6th April and 6th September. This was due to FMAM rains received in the area. After September 6th, the probability of dry spells increased sharply with the cessation of rains. Some farmers’ of this cluster had started planting high yielding medium maturing hybrid maize varieties from the beginning of mid-December despite the prevalent dry spell. This could be due to the higher altitude that had the cooling effect and clay soils that store moisture in the soil for an extended period until the main rainy season comes. However, clay soils also pose the threat of water logging and anaerobic conditions that damage plant stands. Moreover, planting in late December and early January exposed maize seeds to a base temperature of higher than 8°C (cutoff temperature for most processes) and rainfall of less than 50 mm/month (Fig 4) that allowed germination of seeds but coincided with sufficient moisture as of mid-March when maize plants start tasseling and silking (period of maximum water requirement) in 70th and 75th days after planting, respectively.
Bilate cluster–The probability of dry spell greater than 7, 10, and 15 days was above 50% in the 22nd of March, 9th of March, and 20th of February in Bilate cluster, respectively (Fig 5). This indicates that planting maize before the 22nd of March has a failure probability of 50% (every other year). The probability of dry spell length greater than 7, 10, 15, and 25 days was lower than 20% starting 2nd of April, 20th of March, 6th of March, and 9th February, respectively. Dry spell probability of greater than 7 days became closer to 10% starting the second dekad of April. However, risk adverse resource poor farmers delay planting maize until the dry spell probability decreases further in May and exercise planting early maturing maize varieties. The late planting was meant to coincide with the period of optimum soil temperature and reduced evaporation. However, risk taker farmers that had access to irrigation water and conservation practices might start planting early when the dry spell probability was about 50%. The probability of dry spell length greater than 7 days was more than 15% between the 3rd dekad of May and July. This risky period coincided with active vegetative growth, weeding, and urea application. Hence, there is a need to manage the dry spell in the middle of the growing season. The sharply curved parabola of dry spell rises after the 2nd dekad of September signaling the end of rainy season in Bilate cluster. The presence of a dry spells exceeding 10 days in such areas agrees with the findings of Barron [67], who showed a dry spells exceeding 10 days in East Africa varied from 20% to 70% or more depending on the onset of the rainy season.
Hawassa cluster–The analysis of dry spell probability showed that the potential water scarce period was between the 2nd dekad of September and 1st dekad of April (Fig 5). It is unlikely to grow maize crops without continuous irrigation water supply in this period. The probability of dry spell length of 7 and 10 days was less than 50% after the 27th and 13th of March, respectively. This period coincided with the land preparation decisions of the farmers. The probability of dry spell length greater than 7, 10, and 15 days were below 10% after April 20th, March 31st, and March 15th, respectively. Thus, lower than 7 days of dry spell occurred on the 3rd and 4th dekad of April and coincided with the maize planting decision of farmers in Hawassa cluster (e.g. Seraro and Hawassa). However, the dry spells of length greater than 7 days rose above 10% in May and June, which coincided with the period of urea application, weeding, and flowering. Dry spells during planting could be managed through dry planting; however, the dry spells in subsequent stages needed critical agronomic management decisions enhancing the available soil moisture. The probability of dry spell length greater than 7 days was greater than 20% after the second dekad of September signaling the end of the rainy season.
Dilla cluster–The production season for rain-fed maize occurred between mid—April to the end of September. The probability of dry spell length greater than 7, 10, and 15 days was less than 50% after March 18th, March 5th, and April 15th, respectively (Fig 5). Thus farmers prepare their maize fields starting the 3rd dekad of March in Dilla cluster. The probability of dry spell length greater than 7, 10, 15, and 25 days was less than 10% after 7th April, 24th March, March 8th, and February 14th, respectively. The first dekad of April marked the beginning of maize planting in Dilla cluster and it signaled the beginning of the rainy season. Probability of dry spell length of greater than 7 and 10 days increased above 20% beginning the 3rd dekad of May and 2nd dekad of June, respectively. Even so, the period starting the 2nd dekad of May coincided with the period of urea application and weeding which required sufficient moisture for uptake by plants.
The ability of maize plants to survive in May and June under such dry spells might be attributed to the higher humidity of the area that practically reduces the evapotranspiration in the rainy period. This necessitated management of dry spells after the 2nd dekad of May in Dilla cluster as the crop usually suffers extended stress. This is against the reality that Dilla cluster received above 1792 mm every other year and attributed to the fact that such large amount of rain comes half in March–April (FMAM season) and half in June, August and September (JJAS season) leaving February, May, June and July the risky periods. Such dry spells during the growing season had a large impact on maize, and neither annual nor seasonal rainfall did fully explain the challenge of the maize crops in the area. Thus, a few heavy rainfall events lead to an erroneous impression that a growing season is good in Dilla area. This agreed with the work of Usman and Reason [68] in South Africa.
3.8 Trends of dryness and wetness of the area
Forty three percent of the period among the years 1991 to 2020 was classified as wet (inclusive of extremely wet, very wet, and moderately wet); 38.3% of the period classified as dry (moderately dry, severe dry and extreme dry) and 18.3% of the period classified as near normal in areas of Shamana cluster (Fig 6). Thus, the drought hit period is one out of 2.61 years. In Bilate cluster, 16.6, 40, and 43.3% periods are near normal, wet, and dry, respectively, making one out of 2.3 years drought hit in the area. In Hawassa cluster, 50% of the duration was classified as wet, 40% dry, and 10% near normal.
Thus, the drought hit period is one out of 2.5 years in Hawassa cluster. Similarly, in Dilla cluster, 40% of the period between 1991 to 2020 was classified as dry, 41.6% as wet, and only 18.3% was near normal (Fig 6). Thus, the drought hit period is one out of 2.5 years in Dilla cluster. The wet conditions (from moderately wet to extremely wet) were recorded in the years 1996, 2005, 2006, 2011, 2019, and 2020. The driest conditions were reported in the years 1992, 2000, 2015, 2016, and 2009. Thus, agricultural drought has occurred once in two to three years in the study areas. This finding is in harmony with that of Bezabih [69] who reported the central Rift Valley of Ethiopia as the hardest hit region of the country in terms of agricultural drought. Hence, there is a need for improved rainfall forecasts and adapt with proven adaptation and mitigation strategies to reduce the risks of crop failure due to growing dryness, which is also suggested by Gbangou [1] in Ghana for similar drought conditions. The trend of SPI was very highly significant (P<0.0001) in all clusters except Shamana during the last 30 years (Table 5). Thus, Shamana station was less affected by drought. The strongest SPI increment was recorded in Aletawondo (S = 173) and Dilla (S = 133), followed by Wolitasodo (S = 115) and Hawassa (S = 87) (Table 5). This is a manifestation of strong drought events which causes damage to agriculture and water resources. In Hawassa and Wolita Sodo, SPI had shown an increasing trend of 8.5% and 17.8% per decade, respectively. In Halaba and Bilate areas, the increment in SPI was 5.9% and 9.5% every ten years, respectively. Thus, SPI was more severe in lowland areas than the higher and middle agro-ecology. The principal causes for such a continuous upward trends of agricultural drought are increasing land degradation, unwise exploitation of natural resources, and climate change [70], and calls for extensive risk mitigation and adaptation measures in the study area.
3.9 Conditional risks of alternative dates of planting
Shamana cluster—The R-Instat model output reveals that a maize variety having 120 maturity periods and water requirement of 400–500 mm can be grown with a success rate of higher than 50% if planted between 80 to 120 days of the year in Shamana cluster (Fig 7). The model output also shows a success rate of less than 10% for any maize variety in Shamana cluster when grown on the 20th of January. The variety of maize having a 120 day maturity period that needs 400 to 500 mm water has a success chance between 50 and 67% if planted from 80 days of the year (DOY) (March 20) to 120 DOY (April 30) in Shamana cluster (Fig 7). That is, three in five to one in two years. For a variety with 150 days of maturity period and water requirement of 400 to 500 mm, the chance of success is about 32–36%, if planted from March 20 to April 30. This marks the success percentage of one in 2.5 years. Maize varieties with 600 mm water requirement and 150 days of maturity also possess a success rate of about 40% if planted between March 20 and April 30 in Shamana cluster. However, farmers plant varieties (Shone or Limu) that require 500 mm and have maturity duration of 150 days as early as mid-January in Shamana cluster, which is actually two months earlier than the time established by risk analysis. This practice carries a success chance of 10%, which is 1 in 10 years. In fact, farmers prefer to take the risk of early planting every year encouraged by clayey textures that has high water holding capacity, high altitude that reduces evapo-transpirative losses, and a premium markets for early harvest in nearby towns. This could be due to the very nature of soil and climate where light showers which produce about 100 ml of stem flow can provide from 40–50 ml of water in East African maize fields and help to moderate the microclimate of the soil [71]. Thus, maize plants collect light rain by the aerial parts and concentrate it around their base in sufficient quantity and accumulate water nearby to favor deeper storage in periods of non-seasonal rains [71]. Thus, if farmers are informed on the onset date, they plan on when to prepare their land and acquire the necessary inputs.
Bilate cluster—A maize variety having 120 maturity periods and 300 mm water requirement has a success rate of 20%, 10%, 13%, and 7% when planted on the 60th, 80th, 100th, 120th day of the year, respectively in Bilate cluster. A maize variety having 120 maturity periods and 400 mm water requirement has a success rates of 3, 7, 13, and 7% when planted on the 60th, 80th, 100th, 120th day of the year, respectively in Bilate cluster. Planting earlier than the 60th day of the year
or delaying the planting date later than the 120th day of the year reduces the success rate regardless of the maturity period and water requirements (Fig 7). This could be due to the low amount of rainfall that cannot meet the high evapotranspiration requirements in Bilate cluster. The already low success rate discourages farmers to grow maize but farmers continue their trial and error from mid-April to mid-May. The success rate could be increased by growing varieties of maize with smaller than 120 days maturity period in this area.
Hawassa cluster—A maize variety having 120 days of maturity and 400 mm water requirement has a success rate of 3, 30, 63, 40, and 30% if planted on the 40th, 60th, 80th, 100th and 120th day, respectively in Hawassa cluster (Fig 7). A maize variety having 150 days of maturity and water requirement of 400 mm has a success rate of 13, 50, 33, 23, and 20% if planting is carried out on the 40th, 60th, 80th, 100th and 120th day, respectively in Hawassa cluster. Increasing water requirement results in decreasing success rate regardless of changes in planting date and maturity group. In fact, farmers of Hawassa cluster prefer growing maize between the 100th to 120th day of the year when the sandy to sandy loam soils of the area accumulate sufficient moisture despite a lower rate of success compared to March 20th planting. This crop suffers dry spell during weeding and urea application but escapes dry spell during flowering.
Dilla cluster—planted maize variety of 120 day maturity period and 400 or 500 mm water requirement has a success rate of 0.7, 3, 63, 47, and 17% when planted on the 40th, 60th, 80th, 100th and 120th day of the year, respectively (Fig 7). Thus, planting by 80th to 100th day of the year seems adequate for planting a 120 day maturing maize variety in Dilla clusters. For varieties with 150 and 180 days of maturity group, the success rate is hardly over 30 and 20%, respectively, despite reshuffling planting dates or reducing water requirements. This model output fits well with farmers’ practice in Dilla cluster. Thus, in clayey soils of Shamana farmers plant earlier than the date recommended by the model, whereas in sandy to sandy loam soils of Hawassa and Bilate farmers delay planting up to two dekads until the soils get sufficient moisture for germinating seedlings. In both cases, the success rate of risk taker farmers’ is higher than anticipated by the model due to local soil and altitude variations. The risk taker highland farmers of Shamana cluster guarantee sufficient maize green harvest, which is vital for food security in Bilate and Hawassa clusters, places which are yet to plant maize.
3.10 Local area seasonal calendar
Shamana cluster–Traditionally, land preparation is done in December, planting is done in January to take maximum advantage of early rains, weeding is done twice in February and March, urea is applied in March, and harvesting (green and dry) is done in May and June, respectively in Shamana cluster (Table 6). Thus, farmers of this cluster require input and prepare their lands in December. According to R-Instat software output, the rainy season occurs between March to September, while the short dry period occurs in June and July. The pronounced risky period (period of dry spells) occurs in January and February during emergence and vegetative growth due to the late onset of rains. Thus, planting in January is considered extremely risky by the software, which is against farmers’ successful practice. This might be partly due prevalence of light rains in DJF (December-February), interception and concentration of light rains by aerial parts and it around their bases in sufficient quantity, and accumulate water near its base, giving deeper storage than would otherwise result from direct penetration during prolonged dry spells [71]. In fact, the whole year was agriculturally important for growers in Shamana cluster as their land is not left idle throughout the year. They had a double production cycle with maize crop followed by potato or common bean (highlands of Aletawondo, Dilla, Halaba, Shala, Duguna, fango, Shamana, Woteraresa, Damot gale, Seraro) or tef (Jarso, Waera, Ajeba) double crops depending on nature of the soils.
Bilate cluster–Land preparation, planting, weeding, urea application, and harvesting (green and dry) were done in April, May, June, July and August, respectively in Bilate cluster (Table 6). Because this area repeatedly fails to support hybrid maize, farmers are gradually shifting to double crops of short cycle crops like tef (dry planted in March) and potato or common bean (dry planted in July) with minimum tillage so that the crops capture incoming rains during the peak rainy season. In fact, land preparation coincides with SOS and planting of early maturing maize varieties is done after two to three weeks of SOS. This is mainly because the evapotranspiration is still high, and the sandy soils of the rift valley floor are blazingly hot by this time [64,72] and require additional moisture to allow safe germination of seeds. Such delayed planting exercise likely avoids crop failure during germination amid short growing seasons, which can compromise crops productivity [73]. The rainy season occurs between mid-April to May and from July to mid-August showing the shortest rainy season period. The growing season is limited to 4 to 5 months and a single production cycle. Moreover, precipitation in the rainy season is low and erratic with uneven distribution due to rain-shadow effects, and risky periods of about 60 days (in June and August) occur in the middle of the growing season making rainfall a major limiting factor resulting in crop failure [74].
Hawassa cluster–Land preparation is done in March and April, planting is done in April, weeding is done twice in June and July, urea is applied in June, harvesting (green and dry) is done August and September in Hawassa cluster (Table 6). The rainy period prolongs from mid-April to mid-September with a short dry periods of about 45 days in May and June. The risky period coincides with the end of vegetative growth and the beginning of flowering [24]. The land is left bare for about 6.5 months/year during which period no crop grows without irrigation water. Thus, the growing season is limited to 5.5 months and a single production cycle. The sandy soils [64] of the cluster could be considered advantageous as they allow producers to grow maize with one or two plowings without further repeated tillage [75]. Planting coincides with the start of the rainy season (last two decades of April) and carried out by rushing to capture the incoming rains. Earlier planting exposes the crop to wild animal attack (birds, dogs, hyenas, etc) whereas later planting exposes the plants to dry spell and stalk borers. However, intercropping or relay cropping of compatible short maturing dwarf varieties of legumes is possible through careful planting arrangement. Framers of Hawassa cluster rightly avoided early planting dates which could have led to crop failure due to long dry spells occurring shortly after planting. Similar situations were reported by Sivakumar [76] in Western Africa.
Dilla cluster—Land preparation is done in February, planting in March, weeding twice in April and May, urea applied in April, harvesting (green and dry) in July/August in Dilla Cluster. In fact, land is prepared with a few rainy showers obtained in February and planting is done prior to the SOS in the last two dekad of March, which is justified by the humid environment [77] and reddish clay to clay loam soils [64] that reduces evapotranspiration and possesses high water holding capacity, respectively. The cluster had peak of rainy season in April and October, but maize plants suffer moisture stress during mid—June to mid-August, which is a risky period of 60 days. The dry season extends from November to February. The culture of coffee, other perennial crops, and agroforestry systems across the dry season was attributed to the prevalence of higher humidity reducing evapotranspiration and reserves of moisture accumulated in deep soils during the rainy season. In fact farmers of this area often plant longer-season maize varieties during the main rains (high concentration of rainfall in April and May) so that they can continue to grow and mature during short rains.
Cropping season depends not only on the start of the rainy season (which is determined purely from the analysis of daily rainfall data) but also on crop type, water holding capacity of soils, temperature, evapotranspiration, humidity, topography, market demand and risk of biotic/abiotic stress in the area. Thus, a better understanding of climatic, edaphic, topographic, socio-economic and cropping systems required prior changing agricultural practices to adapt to weather variability or climate change. In fact, the dominant factor and driving force of variability in weather-crop calendar is the climate of the area due to which farmers require climate-related information to ensure efficient manipulation of the agricultural calendar where they would be able to match crop sensitivity stages with the available moisture so that the maximum effective rainfall occurs during the development and mid- season stages at a period of high crop water demands.
4. Conclusion
Rainfall records covering a period of thirty years (1991 to 2020) were used with the objectives of studying the variability of rainfall and dry spells while also investigating recent trends in rainfall and SPI on maize growing season. The mean annual rainfall in Bilate, Shamana, Hawassa, and Dilla clusters was 912, 1218, 1017, and 1396 mm, respectively. In fact, there is a greater than 50% probability of getting 894, 1208, 1016, and 1386 mm of rainfall every year in Bilate, Shamana, Hawassa, and Dilla clusters, respectively. Indistinctly, the amount of rain seems adequate for the maize because the crop requires about 500 to 800 mm per growing season. However, the maize crop actually suffers stresses of varying degrees across all clusters due to uneven distribution, dry spells and spatio-temporal variability within the growing seasons. In fact, there was a rain event in ONDJ, FMAM, and JJAS seasons. The risk of planting maize is greater than 50% when planted in 12th March, 15th March, 20th March, and 25th March in Dilla, Shamana, Hawassa, and Bilate areas, respectively. Uniquely, the transitional highland areas of Shamana cluster utilized ONDJ rains for land preparation and planting maize prior two months from SOS, taking the risk unlike other clusters, which was mainly due to highland physiography, showers of ONDJ rains and clayish soils that preserve moisture to sustain the maize growth. In Dilla cluster, agricultural calendar coincides with start of season with 63% success rate from conditional risk of farmers planting date analysis. Model based analysis of conditional risk of farmers planting dates showed a success rate of less than 10, 7 and 40% for commonly grown maize varieties in Shamana, Bilate and Hawassa Dilla clusters, respectively. However, the success rate of risk taker farmers’ at Shamana, Bilate and Hawassa is higher than anticipated by the model. Triangulating the model output with observations at grass root level also showed that agricultural operations pertinent for maize production are delayed by one and two months from start of season in Hawassa and Bilate clusters, respectively the result of which has not been anticipated by model outputs. The delay of agricultural operations in Hawassa and Bilate clusters was mainly meant to avoid the risk of crop failure due to hotter belg season, sandy soils and dry spells.
The contribution of JJAS, ONDJ, and FMAM rains to annual rainfall was 41.3%, 30.9%, and 27.8%, respectively, in the southern central Rift Valley. Thus, the contribution of JJAS rains to annual rainfall was lower than that of the country and southern region average. The coefficient of variability for annual rainfall in Abaro and Abaya was 13.1 and 27.5%, indicating lower variability in high altitude areas of Abaro compared to low altitude areas of Abaya. The probability of end of the season greater than 50% is 18th July, 6thAugust, 20th August, and 27th September in Bilate, Hawassa, Shamana, and Dilla clusters, respectively. Thus, cessation is slower in Dilla compared to other clusters, but cessation is faster in Bilate than other clusters, making Bilate and Hawassa clusters risk prone during grain filling and maturity of maize.
Annual rainfall showed increasing trend in Bilate, Ropi and Hawassa areas. However, there was decreasing trend of annual rains in Dilla (0.421 mm/year) and Halaba (1.67mm/year) areas. Also there was decreasing trend of JJAS rains in Shamana, and ONDJ and FMAM rains in Halaba as manifested by the deviation of annual rainfall from the long term mean. The trends of FMAM (February-May) rains were negative for most clusters and those of December and January were negative and significantly (P<0.05) decreased in Bitena, Bilate, Ropi, Hawassa and Aletawondo. The increasing annual rainfall trends despite decrease FMAM rains could be due to increases of rainfall in one or two pick months in the area, and indicates that relaying only on annual or seasonal rainfall trends have misleading outcomes. This depicts why annual rainfall totals has little implication for growing season planning and justifies why farmers’ were reporting increasing challenges from climate change to grow maize crops. In fact, the risks posed by the high variability of the growing season, intermittent dry spells and increasing agricultural dryness were leading structural shift in climate change adaptation in Bilate cluster faster than other areas.
The start of season had a decreasing trend of 1.82 days/ year in Shamana clusters, as opposed to other areas. But, the duration of rainfall had a decreasing trend of 8.33, 3.33 to 5, 2 to 9.17 and 7.14 days/decade in Shamana, Bilate, Hawassa and Dilla clusters, respectively. Trend analysis of standardized precipitation index also showed 43, 43, 40, and 45% of the period between the years 1991 to 2020 were dry in Shamana, Bilate, Hawassa and Dilla clusters, respectively. The increasingly lower than average rainfall duration combined with late onset and early cessation (Shamana, Bilate, and Hawassa) and increasing dry spells after planting of maize (Bilate and Dilla) posed a high risk for maize production. The study also revealed that maize calendar is different for adjacent clusters in the study area, which would be due to variation in the amount and distribution of rainfall and dry spells across clusters and seasons. This could be attributed to macroscale atmospheric and oceanic circulation systems, flows of atmospheric moisture, and their interaction with local edaphic and topographic features. Analyzing the data across all scales indicated that the Southern Central Rift valley of Ethiopia was facing high variability of rainfall during the maize crop calendar that subsequently causes extended dry spells and agricultural dryness of varying degrees following planting of maize, which necessitates the implementation of adaptation measures. There is also a need to to have real time agro-meteorological advisory fitting for each cluster, and which includes soil water holding capacity and topographic variations in order to enable a realistic estimate of planting windows for maize production in the study area.
Supporting information
S1 Fig.
Rainfall Regimes of the country: A: Bi-modal type-1 rainfall pattern B: Mono-modal rainfall pattern b1—single maxima rainfall with wet period running from February/March to October/November, b2—single maxima rainfall with wet period running from April/May to October/November b3—single maxima rainfall with wet period running from June/July to August/September C: Bi-modal type-2 rainfall pattern EMI: Ethiopian Meteorology Institute EPCC: Ethiopian Panel of Climate Change ET: Ethiopia SNNP: Southern Nations, Nationalities and Peoples.
https://doi.org/10.1371/journal.pclm.0000218.s001
(TIF)
S1 Data. Shape files of base map used in the ArcGIS software for Figs 2 and 3.
https://doi.org/10.1371/journal.pclm.0000218.s002
(XLSX)
Acknowledgments
Acknowledgement is forwarded to Hawassa University College of Agriculture School of Plant and Horticultural Sciences for allowing the senior author to join the PhD program in Agronomy. The authors also gratefully thank EMA, CHIRPS, EMI, SARI and HARC for availability of rainfall station data.
References
- 1. Gbangou T, Ludwig F, van Slobbe E, Greuell W, Kranjac-Berisavljevic G. Rainfall and dry spell occurrence in Ghana: trends and seasonal predictions with a dynamical and statistical model. Theor Appl Climatol. 2020;5:1–17.
- 2. Li X, Meshgi A, Babovic V. Spatio-temporal variation of wet and dry spell characteristics of tropical precipitation in Singapore and its association with ENSO, Int J Climatol.2016; 1–16.
- 3.
Muchow RC, Bellamy JA. Climatic risk in crop production: models and management for the semi-arid tropics and subtropics. Wallingford, UK, CAB International;1991.
- 4. Bekele F, Mosisa N, Terefe D. Analysis of current rainfall variability and trends over Bale Zone, South Eastern highland of Ethiopia. Clim Chan., 2017; 3(12): 889–902.
- 5.
Admassu S. Rainfall Variation and its Effect on Crop Production in Ethiopia, Msc thesis submitted to School of Graduate Studies in Civil Engineering, Addis Ababa University, Ethiopia; 2004.
- 6. Wagaye BA, Antensay M. Rainfall Variability and Trends over Central Ethiopia. Int J Environ Sci Nat Res, 2020; 24(4): 556144.0227.
- 7. Jury MR, Funk C. Climatic trends over Ethiopia: Regional signals and drivers. Int. J. Climatol, 2012; 33:1924–1935.
- 8. Seleshi Y, Zanke U. Recent changes in rainfall and rainy days in Ethiopia. Int J Climatol,2004; 24 (8): 973–983
- 9. Segele ZT, Lamb PJ. Characterization and variability of Kiremt rainy season over Ethiopia. Meteorol. Atmos. Phys., 2005; 89: 153–180.
- 10. Viste E, Korecha D, Sorteberg A. Recent drought and precipitation tendencies in Ethiopia, Theoretical and Applied Climatology, 2013;112(3–4):535–551.
- 11. Degefu MA, Rowell DP, Bewket W. Teleconnections between Ethiopian rainfall variability and global SSTs: observations and methods for model evaluation, Meteorol Atmos Phys., 2017; 129:173–186.
- 12. Belay A, Demissie T, Recha JW, Oludhe C, Osano PM, Olaka LA, et al. Analysis of Climate Variability and Trends in Southern Ethiopia. Climate, 2021; 9(96):1–17.
- 13.
Ethiopian panel on Climate Change. First Assessment Report, WorkingGroup I Physical Science Basis, Published by the Ethiopian Academy of Sciences, Addis Ababa, Ethiopia; 2015.
- 14.
Haile T, Yarotskaya L. Onset and cessation of rains in Ethiopia, NMSA Memo, Addis Ababa, Ethiopia;1988.
- 15. Kebede A, Diekkrüger B, Moges SA. An Assessment of Temperature and Precipitation Change Projections using a Regional and a Global Climate Model for the Baro-Akobo Basin, Nile Basin, Ethiopia. J Earth Sci Clim Change, 2013; 4: 133.
- 16.
Mersha E. Proceedings of the National sensitization workshop on agro-meteorology and GIS,17–18 December 2001, EARO, Addis Ababa, 271pp; 2003.
- 17. Elias B, Yaya D. Trend analysis of seasonal rainfall and temperature pattern in Damota Gale district of Wolaita Zone, Ethiopia. Res Squ., 2020; 8760: 1–16.
- 18. Ezeh CU, Obeta MC, Anyadike RNC. Variations in the sequences of daily rainfall across Nigeria, Arab Geosci.2016; 9 (681): 133–147.
- 19. Mangaraj AK, Sahoo LN, Sukla MK. A Markov chain analysis of daily rainfall occurrence in Western Orissa of India. Reliab Stat Stud, 2013; 6(1): 77–86.
- 20. Barkotulla M. Stochastic Generation of the Occurrence and Amount of Daily Rainfall. Pak J Stat Oper Res. 2010; 6(2).
- 21. Heisey PW, Rubenstein KD. 2015. Using Crop Genetic Resources To Help Agriculture Adapt to Climate Change: Economics and Policy, USDA Economic Research Service Economic Information Bulletin Number 139 April 2015.
- 22. Kingra PK, Gill KK, Singh SP. Wet and dry spell analysis for crop planning in submountainous Punjab using Markov chain approach. Agric Phys, 2013; 13(2): 193–202.
- 23. Beza E, Silva JV, Kooistra L, Reidsma P. Review of yield gap explaining factors and opportunities for alternative data collection approaches. Eur. J. Agron, 2017; 82: 206–222.
- 24. Muluneh GM, Admassu SS, Tamuka WM, Zemede AL, Beruk HS, Aman MT, et al. Conservation Agriculture-Based Maize-Bean Cropping Systems in South Central Ethiopia. J Nat Sci Res, 2020;10(12):1–20.
- 25. Rockstrom J. Resilience building and water demand management for drought mitigation, Phys. Chem. Earth, 2003; 28: 869–877.
- 26. Kassie BT, Rotter RP, Hengsdijk H, Asseng S, Van Ittersum MK, Kahiluoto H, et al. Climate variability and change in the central rift valley of Ethiopia: Challenges for rain fed crop production. Agric Sci., 2014;152(1): 58–74.
- 27. Gebremichael A, Quraishi S, Mamo G. Analysis of seasonal rainfall variability for agricultural water resource management in southern region, Ethiopia. J. Nat. Sci. Res. 2014; 4, 56–79.
- 28. Raes D. Frequency analysis of rainfall data Inter-University Programme in Water Resources Engineering (IUPWARE), Leuven, Belgium;2004.
- 29. Maraun D, Wetterhall F, Ireson AM, Chandler RE, Kendon EJ, Widmann M, et al. Precipitation Downscaling under Climate Change: Recent Developments to Bridge the Gap between Dynamical Models and the End User. Rev Geophys, 2010; 48(3): 1–34.
- 30. Bayable G, Amare G, Alemu G, Gashaw T. Spatiotemporal variability and trends of rainfall and its association with Pacific Ocean Sea surface temperature in West Harerge Zone, Eastern Ethiopia, Environ Syst Res. 2021; 10(7):1–21.
- 31. Tan ML, Samat N, Chan NW, Lee AJ, Li C. Analysis of Precipitation and Temperature Extremes over the Muda River Basin, Malaysia. Water, 2019; 11(283):1–20.
- 32.
Stern R, Parsons D, Stern D, Torgbor F. R-Instat Climatic Guide, University of Reading and IDEMS International, UK; 2021.
- 33. Olatunde AF, Adejoh I. Annual Exceedance Probability and Return Periods of Rainstorms in Lokoja, Internat J of Soc Sci, 2018; 11 (3): 40–46.
- 34. Weibull W. A statistical study of the strength of material. Ing. Vetenskaps Akad.Handl. (Stockholm) Vol. 151, pp. 15; 1939.
- 35. Gitau W. Diagnosis and predictability of intraseasonal characteristics of wet and dry spells over equatorial East Africa. Earth Sciences. Université de Bourgogne;2010.
- 36.
Mamo GD. Using seasonal climate outlook to advise on sorghum production in the Central Rift Valley of Ethiopia. PhD Thesis, University of the Free State, Bloemfontein, South Africa;2005.
- 37. Mugalavaia EM, Kipkorira EC, Raes D, Rao MS. Analysis of rainfall onset, cessation, and length of growing season for western Kenya Agric Forest Meterorol,2008;148:1123–135.
- 38. Gabriel KR, Neumann J. A Markov chain model for daily rainfall occurrence at TelAviv. Q J Roy Meteor Soc.1962; 88:90–95.
- 39.
National Meteorology Service Agency (NMSA). 1996. Climatic and agroclimatic resources of Ethiopia., Meteorol Res Rep Ser., 1: 1, Addis Ababa, Ethiopia.
- 40. Manikandan M, Thiyagarajan G, Bhuvaneswari J, Prabhakaran NK. Wet and dry spell analysis for agricultural crop planning using Markov chain probability model at Bhavanisagar. Internat J, Math and Comp Appl Res. 2017;.7(1), 11–22.
- 41.
McKee TB, Doesken NJ, Kleist J. The relationship of drought frequency and duration to time scale. In: Proceedings of the 8th Conference on Appl. Climatol., 1993;179–184.
- 42. Yue S, Pilon P, Cavadias G. Power of the Mann–Kendall and Spearman’s rho tests for detecting monotonic trends in hydrological series. J. Hydrol, 2002;259: 254–271.
- 43. Kendall MG. Rank correlation methods, 4th edition, Charles Griffin, London;1975.
- 44. Daniel FB, Vanessa CL, Armstrong TW. Interpreting Kendall’s-Tau and Tau-U for single-case experimental designs, Cogent Psych, 2018; 5:1–26.
- 45. Sen PKJ. Estimates of the regression coefficient based on Kendall’s tau. J. Am. Stat. Assoc., 1968;63: 1379–1389.
- 46.
Baijukya F, Sabula L, Mruma S, Mzee F, Mtoka E, Masigo J, et al. Maize production manual for smallholder farmers in Tanzania. Ibadan, Nigeria: IITA; 2020.
- 47. FEWS NET (Famine Early Warning Systems Network). Remote Monitoring Report, Tanzania Seasonal Calendar, Typical Year Vuli and Msimu seasons, https://fews.net/east-Africa/Tanzania; 2017.
- 48.
Mesay A. The Onset, Cessation and Dry Spells of the Small Rainy Season (Belg) of Ethiopia. National Meteorological Agency of Ethiopia, Addis Ababa. 10-13pp;2006.
- 49. Wodaje GG, Eshetu Z, Argaw M. Temporal and spatial variability of rainfall distribution and evapotranspiration across altitudinal gradient in the Bilate River Watershed, Southern Ethiopia, Afr. J. Environ. Sci. Technol, 2016;10(6): 167–180.
- 50. Hachigonta S, Reason CJC, Tadross M. An analysis of onset date and rainy season duration over Zambia, Theor. Appl. Climatol. 2008;.91: 229–243.
- 51.
Natai C. Seasonal climate variability analysis and characterization for major crops and vegetables production in Siraro-kKofele potato-vegetable livelihood zone of Ethiopia. MSc Thesis, Haramaya University, Ethiopia;2016.
- 52. Bahiru W, Mekoya A. Rainfall Variability and Trends over Central Ethiopia, Int J Environ Sci Nat Res. 2020; 24(4):1–13.
- 53. Thompson K, Chidawanyika F, Kruszewska I, Tirado R. Building resilience in East African agriculture in response to climate change. Greenpeace Research Laboratories Technical Report: 05;2015.
- 54. Takele R, Tesfaye K, Traore PCS. Seasonal Climate Predictability in Ethiopia. Identifying the best predictor sets for sub-seasonal to seasonal forecasting. CCAFS Working Paper no. 301. Wageningen, the Netherlands; 2020.
- 55. Gissila T, Black E, Grimes DIF, Slingo JM. Seasonal forecasting of the Ethiopian summer rains. Int. J. Climatol.2004; 24:1345–1358.
- 56. Korecha D, Barnston AG. Predictability of June-September rainfall in Ethiopia. Mon. Weather Rev, 2007;135: 628–650.
- 57.
Krauer J. Rainfall Erosivity and Isorodent Map of Ethiopia. Soil Conservation Research Project Report 15, University of Berne, Berne, Switzerland;1988.
- 58.
Beyene EG. Geospatial Analysis to Study Environmental Change: Climate Variability and Vegetation Cover Dynamics in Ethiopia and the Horn of Africa, PhD Desertion, University of Berlin, Germany; 2010.
- 59. Gleixner S, Keenlyside NS, Demissie TD, Counillon F, Wang Y, Viste E. Seasonal predictability of Kiremt rainfall in coupled general circulation models. Environ. Res. Lett, 2017;12: 114016.
- 60. Befikadu E, Simane B, Teferi E, Ongoma V, Tefera N. Climate Variability and Farmers’ Perception in Southern Ethiopia, Adv Meteor., 2019; 7341465: 1–19.
- 61. Hessebo MT, Woldeamanuel T, Tadesse M. Spatial and temporal climate variability and change in the Bilate catchment, central Rift Valley lakes region, Ethiopia, Phys Geogr., 2019;1698094: 1–28.
- 62. Cheung WH, Senay GB, Sing A. Trends and spatial distribution of annual and seasonal rainfall in Ethiopia. Int J Climatol., 2008; 28: 1723–1734.
- 63. Viste E, Korecha D, Sorteberg A. Recent drought and precipitation tendencies in Ethiopia, Theoretical and Applied Climatology,;2013; 112(3–4):535–551.
- 64. Makin J, Kingham TJ, Waddams AE, Birchall CJ, Teferra T. Development prospects in the Southern Rift Valley, Ethiopia. Land Resource Study 21, Land Resources Division, Ministry of Overseas Development Tolworth Tower, Surbiton, Surrey, KT6 7DY, England;1975.
- 65.
Bank World. Ethiopia: Managing water resources to maximise sustainable growth. Washington, D.C. World Bank;2006.
- 66. Ayele K. Analysis of Trend and Variability of Climate of Hawassa Based on Surface Observations, Int J Clim Res. 2020; 4(1): 16–23.
- 67. Belay A, Recha JW, Woldeamanuel T, Morton J F. Smallholder farmers’ adaptation to climate change and determinants of their adaptation decisions in the Central Rift Valley of Ethiopia, Agric & Food Secur. 2017; 6 (24):1–13.
- 68. Usman MT, Reason C. Dry spell frequencies and their variability over southern Africa. Clim Res, 2004;26:199–211.
- 69. Bezabih E, Hadera G, Nigatu R. Impacts of Improved Seeds and Agrochemicals on Food Security and Environment in the Rift Valley of Ethiopia: Implications for the Application of an African Green Revolution: Dry lands Coordination Group Report No. 56 02, 2010, DCG c/o, MiljøhusetG9. Norway. Oslo.
- 70. Thomas E, Jordan E, Linden K, Mogesse B, Hailu T, Jirma H,et al. Reducing drought emergencies in the Horn of Africa. Sci Total Environ.,2020;727: 138772. pmid:32339831
- 71. Glover J, Gwynne MD. Light Rainfall and Plant Survival in East Africa I. Maize: J. Ecol., 2016; 50 (1): 111–118.
- 72. Gisha HK, Fufa FF, Adugna TD. Hydrologic responses to climate and land-use/land-cover changes in the Bilate catchment, Southern Ethiopia. J Wat Clim Chang, 2021;12: 8, 3750.
- 73. Laux P, Kunstmann H, Bárdossy A. Predicting the Regional Onset of the Rainy Season in West Africa; International Journal of Climatology, 2008; 28 (3):329–342.
- 74.
Improving Productivity and Market Success (IPMS). Alaba Pilot Learning Site diagnosis and program design, Addis Ababa, Ethiopia;2005.
- 75. Beshir B, Birhanu TA, Wegary D, Mekuria M, Merga FL, Mupangwa W, et al. Achievements and prospects of CASI practices among smallholder maize–legume farmers in Ethiopia, In: SIMILESA, pp.227 – 262; 2020.
- 76. Sivakumar MVK. Predicting rainy season potential from the onsetof rains in Southern Sahelian and Sudanian climatic zones of WestAfrica. Agric For Meteorol, 1988; 42(4):295–305.
- 77. Shengu MK. Seed System and Quality Estimation of Maize (Zea Mays L) in Humid Tropics of Southern Ethiopia. Intern J of developing Res, 2017;7 (3):12057–12073.