Agricultural development is fueled by economic growth, often ignoring the environmental degradation that it brings (den Besten et al., 2016; Pande and Savenije, 2016; Roobavannan et al., 2018). Due to the coevolution of human well-being with its agricultural water systems (Montanari, 2015; Roobavannan et al., 2017a), understanding the underlying interlinkages between the two is of utmost importance for sustainable agricultural development (Sivapalan et al., 2012; Montanari, 2015; Roobavannan et al., 2017b). One key challenge is the complexity induced by competing water demands by various users such as upstream vs. downstream water users, large vs. small holding farmers and the farmers vs. the environment (Gober and Wheater, 2014). This competition adversely affects those whose demands either have hidden or low economic value and have little or no influence on water resource policy, such as downstream users, smallholder farmers and the environment.

On the other hand, rainfall variability and other factors such as volatility in market prices drive farmers to choose the types of crops they grow (Ruben et al., 2000; Niles and Brown, 2017). Rain-fed agriculture is common in sub-Saharan African countries, mostly practiced by smallholder farmers (Bank, 1997; Strzepek and McCluskey, 2007), where crop yields depend on the strength of the rainfall seasons (Rockström et al., 2003; Jemberie et al., 2016; Villani et al., 2018). The income of smallholders is therefore linked to low productivity of crops grown and market volatility, and such dependence in turn results in low incomes for such farmers.

Ethiopia depends on rain-fed agriculture with limited use of irrigation for agricultural production (Awulachew et al., 2010). It is estimated that more than 90% of the food supply in the country comes from low productivity rain-fed smallholder agriculture, and hence rainfall is the single most important determinant of food supply and the country's economy (Mati, 2006; Manaswi and Thawait, 2014; Zewdie et al., 2020). Gebre et al. (2015) investigated the effects of climate on water availability in the Dhidhessa River basin and noted the effects of spatial and temporal variations in rainfall on agricultural production. Adgolign et al. (2015) reported that for effective planning and management of water resources, understanding the spatial and temporal fluctuations of water flows is critical. The authors recommended that a simulation approach might be the best approach to quantify water availability. However, there is limited literature on the dependence of farmers on rain-fed agriculture, effects of population, local re-adaptation to climate change variability and management in the Dhidhessa River Basin.

If such farmers mostly depend on rainfall, the key is to understand how they adapt to its variability. Here migration is often suggested as one adaptive response to climate variability (Andersson, 2014; Teweldebrihan et al., 2020). Yet smallholders are known to be efficient and resilient producers, with novel traditions to sustain their livelihoods under adversity (Debela et al., 2015; Belay et al., 2017; Dechassa et al., 2020).

This paper assesses how smallholder farmers adapt to anthropogenic and climate induced variability in the Dhidhessa river basin. In order to do so, the paper (i) interprets the evolution of crop coverage over time in the basin and (ii) understands how crop production and patterns depend on rainfall in the basin.

The Dhidhessa River is a tributary of the Abbay River emanating from an elevation above 2,500 m near the Wacha and Vennio mountainous terrains in Ethiopia with a basin area of 16,567 km² (Figure 1).

The climate of the Dhidhessa basin is traditionally characterized mostly as a Woina Dega (Sub-tropical) and Kolla (tropical) with heavy rainfall during Kiremt (winter) season (Bekele et al., 2021). The basin experiences variable rainfall, which ranges from a minimum of 121 mm to a maximum of 2,199 mm annually. Average rainfall amounts for the duration of dry, short and long rainy seasons are 151, 218, 916 mm, respectively (Tesemma et al., 2010). Overall annual rainfall decreases from the South-West (over 2,000 mm/year) to the North-East (about 1,000 mm/year), and about 70% of the rain falls between June and September (Conway, 2000; Dechassa et al., 2020).

The elevations in the basin are classified as highland (>2,500 m a.s.l), midland (1,500–2,500 m a.s.l) and lowland ( <1,500 m a.s.l.) (Chimdessa et al., 2019). The temperature in the basin fluctuates depending on the topography of the sub-catchment area. The minimum average temperature in the sub-basin ranges between 7 and 17°C, and the maximum average temperature ranges between 21 and 37°C (Chimdessa et al., 2019; Tolessa et al., 2020). Studies made in recent years indicate that climate change, land and water degradation are mainly related to rainfall fluctuations and the traditional beliefs dominating in the community toward water (Adgolign et al., 2016; Chimdessa et al., 2019).

The results demonstrate that farmer decision to grow which crops are sensitive to rainfall and their expectations of crop prices resulting from rainfall variability. The farmers' decisions and crop prices are endogenous in how cropland areas do not respond to rainfall variations, but instead, farmers shift from one crop to another. This demonstrates the sociohydrological complexity of the smallholder farming system (Pande and Savenije, 2016; Niles and Brown, 2017; Jovanovic et al., 2020; Lyu et al., 2020a; Pande et al., 2020). The interviews corroborate that the farmers decide on the types of crops to grow every season based on their prediction of rainfall using traditional rainfall prediction practices (Balehegn et al., 2019; Wedajo et al., 2019).

The interviews suggest that more farmers grow Teff in years when the expected amount of rainfall is low using the indigenous (traditional) knowledge in which farmers predict the supply will decrease and the price will rise due the demand of Teff consumers in the country. The farmers in the area mostly decide the type of crop to grow using the traditional expectation and past experience, therefore, they used to say “expecting is better than what you have gotten.” This is also corroborated by secondary data, as shown in Figure 9 above. Hence, the ratio of teff grower vs. maize and sorghum in relation to the annual rain fall varies in which the ratio of HH number grow teff vs. maize shows < 1 and ratio of teff grower vs. sorghum is about >1, respectively. Consequently, the less number of households start to grow Teff in good rainfall years (it does not matter whether Teff growers are more or less than Maize growers). This leads to an increase in the per-unit cost of producing Teff relative to other crops due to the rise of cultivation demand. The cost for teff is higher that maize and sorghum because the row sowing of teff cultivation needs more time and intensive labor cost (row-making and weeding due to its grain size) than other crops. This can be observed in Figure 10 that plots the per unit cultivation cost of Teff vs. other crops. Consequently, the price of seeds, human labor, and related costs of growing Teff is greater than of other crops. Yet these costs are offset by the higher price of Teff under poor rainfall conditions relative to the other two dominant crops, still making Teff an attractive option to grow as suggested. Indeed as Figure 11 shows, Teff prices relative to other crops are higher for seasons with lower rainfall.

About 65% of interviewed farmers said they grow crops that fetch high prices when the rainfall is low. Only about 35% of interviewed farmers suggest that they grow crops when higher rains are expected. Others said they leave their lands for grazing during high rainfall seasons, driven by the expectation that livestock production income will be better than agricultural production (~51% respondents). This is also suggestive of the role that past experiences play (Baldassarre et al., 2013; Viglione et al., 2014; Fuchs et al., 2017; Leong, 2018) in farmers' decision-making to expect higher production levels and, therefore, lower prices for crops, during high rainfall seasons.

The interviews reveal that 81% of farmers make decisions, such as which crop to grow, based on indigenous knowledge. The remaining 19% use the advice from the agricultural extension experts. This may be the reason why the basin's farm system is sensitive to rainfall, but rainfall effects are endogenous. Farmers adapt their cropping patterns based on rainfall and price expectations, indicating that such adaptive measures are in place to make them more resilient to climate variability.

The adaptation strategy also enables them not to respond to climate variability by migrating out of the basin. Despite rainfall variability, households engaged in crop production have steadily increased in the basin. The total population number was 6,072,485 in 2001 and 12,884,927 in 2020 and the total number of growers increased from 1,214,497 in 2001 to 3,681,408 in 2010. For example the number of Teff growers in the basin was 375,003 in 2001 and increased to 1,028,804 in 2019 which is more than double in the time period. Similar increases are observed for other crops. This evidence of the critical role played by adaptation to climate variability is counterfactual to a pattern often alluded to that migration is the only response to climate variability. Lyu et al. (2020b) similarly found no links between rural to urban migration and climate variability in Jiangsu province China. Other studies (Akay et al., 2012; Hagedorn et al., 2013; Delango, 2019; Lyu et al., 2019) have made similar observations, though studies such as Heitmueller (2005) and Roobavannan et al. (2017a) have highlighted it as a response to change in water policy and consequent unemployment.

Such a response to climate variability indicates the extent to which crop production in the basin depends on rain and not so much on the river flows. Traditional practices have perhaps emerged due to long-term dependence on rainfall and its endogenous effects (Simelton et al., 2009; Panda et al., 2013; Mekuria, 2018). The number of households growing crops has been consistently increasing over time without any sensitivity to rain.

This paper investigated how smallholders adapt to the change in climate and climate variability in the Dhidhessa river basin. Analysis of secondary and primary data on land cover, households engaged in crop production, crop yields, crop prices and area covered by three major crops (Teff, maize and sorghum) over a period of 20 years revealed that irrespective of rainfall variability, more farmers are moving in the basin because the area under crop cultivation, as well as the number of households practicing it, has grown over the years. Farmers' decisions to grow which crops were found to be sensitive to rainfall and their expectations of crop prices resulting from rainfall variability and climate change. Results from the trend indicate that the farmers are resilient to the change in climate and farmer decisions in choosing the type of crop respond to the rainfall variability and climate change over the years. Yields were found to be sensitive to rainfall to such an extent that even when fewer farmers grew crops in good rainfall years, the production was higher in spite of a smaller area under cultivation. Further, farmers grow more of Teff relative to other crops in lower rainfall years due to the lower sensitivity of Teff prices to rainfall variability and farmers' expectations of higher Teff prices relative to other crops in lower rainfall seasons. The behavior thus unraveled offers a way forward to incorporating adaptation mechanisms in sociohydrological models to simulate water futures in similar basins worldwide realistically.

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.

The initial concept writing, analysis, review, and final write-up was done by MT and SP in which HL contributed on the analysis. MM contributed on the conceptualization and review of the paper. All authors contributed to the article and approved the submitted version.

The research grant from the International Foundation for Science (IFS) is acknowledged for the financial support (Grant Number I-2-W-6357-1). The support of IHE Delft and Delft University of Technology is acknowledged. The Schlumberger Foundation, Faculty of the Future, is warmly recognized for providing financial support.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.