In the Great Ruaha River Basin of Tanzania, climate stress in agriculture is significantly about water, which the government seeks to address through irrigation. However, a neglected tradeoff in this context is that the extreme weather episodes in the last decades have adversely affected not only rainfed farming but also irrigation due to recurrent and intensified river droughts in this basin. The years of droughts coincide with the years of below-average rainfall driven by El Niño and La Niña weather events. Based on the Tanzania Meteorological Agency (TMA) data (2016), the basin region (the Mbeya Region) has experienced new rainfall extremes in 1986, 1993, 1998, 2005, 2011, and 2015 as Figure 1 shows (TMA Tanzania Meteorological Agency, 2021). The El Niño Southern Oscillation prompted prolonged droughts in 1992–93 and 1997–98 (Kijazi and Reason, 2009). From 1998 until 2005, most parts of the country, including the Great Ruaha Basin, experienced at least two consecutive droughts, with delayed rainfall onsets and unevenly distributed and below-average rains (Kijazi and Reason, 2009). In 2014–15, another drought was caused by “the worst El-Niño” until then, as the FAO (2016: 1) reported, which extended into late 2016, with devastating consequences for livelihoods.

The climate data excerpted from the TMA database is consistent with farmers’ perceptions of historical rainfall changes. The interviewed farmers argued that since the first paddy irrigation in 1998 (when the irrigation scheme was constructed), rains have been falling with more extended delays, with less precipitation, and ceasing earlier than usual. Especially the 2015–16 drought was felt strongly: as the annual report compiled by the scheme authorities confirmed, “the 2015–16 rainfall crisis severely affected farmers in the village and caused crop failures” (Annual Report 2016/17). Farmers contended that many of them had encountered crop failures this year, some already at the beginning of this season, due to the delayed rainfall begin; the seedlings had desiccated before sprouting and yielding crops. Consequently, they could not generate sufficient income to afford basic livelihoods needs, such as food and travel. These dramatic outcomes influenced the National Food Reserve Agency and the World Food Program to plan emergency food purchases in the region to secure rural income and store rice (staple food) in national warehouses to prepare for potential food shortages.

Despite the insights pointing to a strong connection between climate events and local rainfall extremes, Tanzanian policies have mainly focused on the potential role of the growing agriculturalist population in exacerbating water stress in river basins and sought to address this. This misplaced focus was to some extent driven by increasing water competition between water users, leading them to lobby for expulsion of smallholders out of the Great Ruaha Basin. Two decisive events that escalated this discourse were, first, the river’s disappearing water flow for the first time in 1993 and then, the Mtera and Kidatu dams (meet for more than half of the country’s power demand) entirely drying out in 1998, which caused days-long blackouts and paralyzing impacts on the economy (SMUWC, 2001). Scholars found that the local farming, hunting, and tourism businesses alleged small farmers as the main cause of the recurrent droughts in 1998–2005 (Lankford et al., 2009; Walsh, 2007; Walsh, 2012). Hydrological studies did not confirm the veracity of these allegations: the UK-funded project called Sustainable Management of the Usangu Wetland and its Catchment (SMUWC), one of the few and the most informative accounts on the hydrological change in this basin, listed agricultural water use as one of the potential drivers of the sporadically receding river flows, without making a firm statement (SMUWC, 2001). Nevertheless, in 2005, upon his election, president Kikwete acknowledged these claims, ordering “immediate urgent action” to revert the water crisis, as Walsh (2012) quoted, which included conservation area expansions to repel smallholders off the basin. Neither this decision nor studies examining this political discourse paid attention to the potential impact of the 1998–2005 climate events on this hydrological change.

The case study (the Madibira irrigation scheme) as a large-scale smallholder irrigation project played a critical role at the heart of this basin water debate, pulling hostility from the local business circles. The scheme construction was approved in the early 1990s when the government embarked on irrigation expansions to foster rural food and income security (SMUWC, 2001). The scheme abstracts water from the Ndembera River, one of the three major tributaries of the Great Ruaha (along with the Kimbi River and the main branch of the Great Ruaha River) before it flows into the Ruaha National Park (Figure 2). Its construction started in 1993 and ended in 1998, coinciding with the years of the two most dramatic river droughts, inevitably accentuating its potential role on the river droughts. At the end of the 1998–2005 droughts, a paper drafted by a local tourism company addressed this scheme as one of the causes of the droughts and accused the African Development Bank of funding this scheme without an environmental impact assessment and bringing “huge number of migrant people associated with rice farming” to this area (Fox, 2004: 4). Scholars claimed that this paper was backed by the state-owned Tanzania Electric Supply Company, which manages the dams, accusing the scheme and farmers of the desiccated river and the countrywide electricity crisis, allegedly attempting to veil its own water mismanagement of the dams (Machibya et al., 2003; Yawson et al., 2003; McCartney et al., 2007). This paper was influential upon the president’s approval of a game reserve expansion in 2006, followed by the national park expansion encircling the reserve area in 2008 and the banning of agriculture therein (Walsh, 2012). The military pitched camps to evict agrarian groups and burned homes and sheds to ensure that farmers left and herders with their 300,000 cattle moved to the coastal region designated for them (IWGIA International Work Group for Indigenous Affairs, 2016). Two of the six dismantled villages (Mapogoro and Ikoga) were relocated to Madibira, folding its farmer population.

Interviews with farmers and the village chairpeople showed that the increasing farmer population due to this inter-district migration led the land and water demand in the Madibira irrigation scheme to rise instantly. But at the same time, with the Water Resources Management Act of 2009, the government enacted a formal water rights system to restrict water access in irrigation schemes, not only uncontrolled small-scale water use outside them. Water rights were first introduced in 1993 following the river drought this year to promote wise water use. Rights as statutory titles indicated the water volumes that water users are authorized to abstract from rivers, while fees, calculated based on these authorized volumes, had to be paid by them (Koppen et al., 2004). However, scholars found that the outcome of this experiment was unsatisfactory: basin officials lacked administrative capacities to monitor and sanction abstractions according to rights, and the system did not prevent the Great Ruaha from periodically drying and causing water scarcity (Koppen et al., 2004; Maganga et al., 2004). Still, with the 2009 act, the government renamed water rights as water permits and permanently passed them to restrict irrigation within irrigation schemes, in addition to scaling down agricultural land and water use outside them, and prioritized the Great Ruaha’s flows into other sectors, though irrigation has also been a policy priority for decades.

Though the permits are not updated frequently, interviews showed that basin authorities periodically cut off or decreased irrigation water access during dry spells (when farmers need water the most) and continually increased the permit fees. The Madibira irrigation scheme received its final permit in 2013, allowing water abstraction from the Ndembera River from November to October every year. As the scheme’s irrigation officer reported, the permit fee gradually increased from TSH7 to 11 million until 2016, and further changes followed the drought this year (Key informant 1, December 12, 2016). On the one hand, in 2016, basin authorities ordered him to halve the scheme’s water abstraction from 7.5 to 4 cubic meters per second to retain water in the riverbed, although the permit fees would further rise in the same year. The Annual Report 2016/17 supports this statement that decreasing the scheme’s water abstraction is necessary “to restore the perennial water flow of the Great Ruaha River, which will be prioritized.” On the other hand, additionally to halving water abstractions, from 2017 onward, the scheme received water only from January to July, with a five-month cut in the irrigation period. However, as farmers indicated, paddy cultivation traditionally requires a longer time, from December to July, from seedling preparation until harvesting. The late cultivation by one month in January means potential harvesting delays into August, when farmers needed but lacked water due to the end of irrigation and cessation of rainfall, both in June. Together with the increased permit fee, water restrictions heightened the risk of crop loss and strained many livelihoods.

These findings show that irrigation as a climate adaptation strategy has limits in practice. Though the Tanzanian government has hailed irrigation for improved food and income security since the 1990s, it also took contradictory steps to restrict it, bringing about land and livelihood losses. This threat retains its significance since the government neither recognizes climate impact on natural resources nor the agricultural and nonagricultural sector interests standing in conflict in the context of the ongoing basin debate that is political in nature. This omission countervails the prospect of expanded irrigation to foster adaptation for small farmers. Another contradiction in this context is that while the water stress has been blamed on farmers, a narrative which drove conservation area expansions and strict water regulations, in another context promoting the shift to intensive farming, this stress has been considered a consequence of climate change. Though they themselves cut down the Madibira irrigation scheme’s water access, backing the claims that farmers used too much water, two authorities argued that the rising water stress resulted from “global warming” but would be “compensated through efficient commercial agriculture,” with the use of the drought-tolerant hybrid seeds (Key informants 2 and 3, December 13, 2016). Various potential reasons for water stress being reworked in different contexts is another political insight that points to limits of irrigation as adaptation. The next section explores whether the shift to intensive paddy farming has indeed been rewarding in this context.

In SAGCOT, intensive paddy farming with drought-tolerant seeds is promoted as a way to foster climate adaptation and food security. Though some multinational SAGCOT partners, such as Monsanto and Syngenta, sell genetically modified seeds across the corridor, specifically in the rice subsector in the Mbarali cluster, most drought-tolerant rice seeds are bred by the national agricultural research institutes that have received donor support since the 1980s. These seeds are hybrid varieties, combining productive and morphological qualities of traditional and genetically modified seeds, and advertised by the government and partners to sustain and increase crop yield with less water and in a shorter time than usual. In addition to their productivity features, some offer an enhanced taste, which contributed to the widespread acceptance of some varieties by Mbarali farmers since 2013. The most successful hybrid rice seed combining high yield with semi-aromatic flavor is SARO5 (termed TXD306), bred by the Agricultural Research Institute Chollima-Dakawa, with TXD (Tanzania Cross Dakawa) indicating its origins. Other hybrids, such as Katrin (IR54 and IR64), bred by the Kilombero Agricultural Research Institute (KATRIN) in Morogoro, are more productive than SARO5 but did not succeed due to their lack of aroma (most farmers produce rice for both cash and household consumption). Still, farmers locally call SARO5 “the export variety” since it is highly demanded in export markets and brings cash through mass production. Meanwhile, the traditional variety is preferred mostly when taste and domestic market supply are farmers’ priorities.

In Madibira, SARO5 has succeeded, with an above 90 percent acceptance rate by commercial farmers. Various informants (farmers, local authorities, and companies) confirmed that this variety indeed yields more paddy. Moreover, though it requires irrigation for such high-yielding performance, SARO5 crops mature faster, enabling adaptation to shortened irrigation and rainfall periods, for which they are considered drought-resistant. The government extension officers argued that SARO5 seeds can multiply the average paddy production under irrigation from the country average of 4–5 tons per hectare with traditional seeds to 8–9 under intensive farming and 12 under systematic rice intensification (SRI) (kilimo shadidi) (USDA United States Department of Agriculture, 2018). SRI is an intensive rice farming technique developed in Madagascar in the 1980s, which, under some conditions, can provide two times the yield with half of the amount of water required by some traditional seeds (Cornell University, 2020). Spreading the SRI practice along with expanded irrigation to increase paddy production from the countrywide traditional rainfed average of 1–2 tons per hectare is one of SAGCOT’s subsectoral goals (SAGCOT, 2011a). Though progress has been slow: in Madibira, in 2017, only ten farmers practiced SRI, as confirmed by the extension agents. The interviewed farmers considered it a labor-intensive technique needing to count seeds and sow according to a linear pattern (instead of randomly planting in the field) and flatten the soil for which they lacked capital. Still, without SRI, SARO5 has fulfilled its increased production and decreased water requirement promises to some extent, which contributed to their widespread acceptance in a short time. As a farmer put it, “after seeing their neighbors harvest a lot and prosper, even those skeptical of this seed at first only cultivate it today” (Farmer 1, December 04, 2016).

The problem about the SARO5 seeds little mentioned in the scholarly and gray literature is that despite their attractive features such as reasonable price, less crop water requirement, and higher productivity than the local seeds, they need intensive fertilizer use. Tanzania does not have a strong fertilizer industry and imports fertilizers, making their intensive use expensive for smallholders. Therefore, fertilizer use across Tanzania has been low, only one to two bags per hectare to none (Majule et al., 2014). In comparison, Madibira farmers used three-four bags per hectare before shifting to hybrid seeds and seven and more after. Moreover, they used YARA fertilizers only. YARA is a Norwegian fertilizer company; it built a fertilizer terminal at the Port of Dar es Salaam shortly after becoming an executive SAGCOT partner and monopolized the fertilizer market by opening retail stores in rural areas. In Madibira, YARA salespeople were more active than the government-hired local extension agents; they frequently visited villages, organized farming workshops, and staged games with free fertilizers for the winners, successfully promoting own products. Also, the extension agents advised farmers to use only YARA fertilizers due to their “higher quality” (Key informant 4, July 13, 2017) and promoted the need to use at least seven bags of this brand fertilizers in the extension curriculum. Though farmers agreed to increase fertilizer use to sow SARO5 for increased production, production costs climbed as well: in the 2016–17 season, one bag cost TSH55,000 ($24) at the village salespoint, TSH40,000 ($17) directly from YARA, and TSH65–67,000 ($28–30) from intermediaries, all of which farmers considered expensive ¹ .

The shift to intensive farming with SARO5 seeds and increased fertilizer use increased production costs by 50 percent from TSH1.5 to TSH2.5 million ($1,100) per hectare on average, which was beyond the capacity of the most low-income smallholders to afford. This gave rise to loan dependence in Madibira: 60 percent of my informants were smallholders (with one-hectare landholding), and all of them had to rely on loans to engage in paddy cultivation. Though national banks, such as the National Microfinance Bank and Cooperative Rural Development Bank, have joined SAGCOT as partners to give out loans to smallholders, in the widespread lack of collateral to put up for the loan, smallholders in Madibira lacked access to them. The increased production costs thus worsened their dependence on local microfinance institutes and moneylenders (usually, the wealthy local settlers) who charged high interest rates, often exceeding 30 percent. Moreover, some moneylenders used strong-arm tactics to collect loan payments from debtors with low harvest and crop failure by seizing their crops, lands, and small properties at the end of cultivation seasons. Hence, by encouraging the smallholders, particularly the low-income ones, to intensify paddy farming and bear increased costs before establishing a robust microfinancing system, SAGCOT put smallholders at debt risk.

Interviews also cast light on an economic-environmental tradeoff in the shift to intensive paddy farming: farmers stated that they felt “forced to use fertilizers intensively” since the soil fertility steeply declined and soil salinization became a serious problem due to monocropping (Farmer 2, December 05, 2016). They argued that in the first years of irrigation in the scheme in the early 2000s, they harvested 4–5 tons per hectare of paddy without any inorganic fertilizers, which is allegedly nowadays impossible to obtain without multiple bags of fertilizers. Farmers added that intensified droughts accelerated soil salinization, exacerbating the impact of erratic weather on food production. Some findings in the literature align with this suggestion that droughts worsen soil salinity, while salt, in turn, prohibits water uptake by plants (Paul et al., 2019; Corwin, 2020). To address this problem, Tanzanian research institutes developed a salt-tolerant version of the SARO variety, called SATO, which is becoming a necessity for farmers to be able to sustain food production. These findings show that while attempting to nurture food and income security through a shift to intensive farming, the government and SAGCOT partners neglect the rising environmental and economic disadvantages that consequently put greater stress on livelihoods instead of alleviating it.

Whether intensified irrigated paddy farming is a rewarding adaptation strategy for smallholders to sustain and improve their livelihoods requires examining the changing harvest and income levels. This section sheds light on this change following the shifts to irrigated paddy farming in 1998 and to intensified irrigated paddy farming in 2013 in Madibira, focusing on the 2016–17 season, the fieldwork period.

In Madibira, paddy irrigation began with equitably allocated landholding to local subsistence-oriented farmers, one hectare per capita, to commercialize the existing paddy farming. With this size of land, the government sought to put smallholders, who in the region typically held only a few acres (one hectare is about 2.5 acres) (Franks et al., 2013), at the threshold of transition to middle-scale farming and raise their incomes. Interviews showed that though most landholders initially lacked capital and labor to farm such a scale without loans, their incomes and cultivation capacities gradually improved. The scheme office recorded that the scheme-level harvest average was 4–5 tons per hectare in 2004, increased from 1 to 2 tons per hectare before irrigation begun in 1998, the last year of the official harvest records kept by this office. During these years, an assessment study conducted by the government showed that average income rose from TSH145,000 to TSH360–400,000 per hectare in the same period (AFD African Development Fund, 2004). This study also pointed to livelihood improvements associated with harvest and income increases, measured by increased ownership of burnt brick houses, power generators, motorcycles, farming and milling machines, and village shops. Moreover, school enrollment rates rose, and farmers began hiring laborers, prospering, and creating jobs simultaneously (AFD African Development Fund, 2004). The scheme office argued that this scheme-level harvest average of 4–5 tons was maintained until the transition to intensive paddy farming in 2013. However, interviews revealed that some farmers prospered more than others; the low-income farmers that SAGCOT sought to benefit struggled and often failed to achieve and sustain this harvest level. This finding builds on another one that scheme farmers in practice held lands of diverse sizes—small, middle, and large-scale lands—despite the formal landholding rule requiring them to hold a maximum of one hectare per capita. And mainly, the middle- and large-scale farmers harvested a lot, thus pulling the scheme-wide paddy harvest average above the average of the smallholder subgroup, which this study focuses on.

Despite holding one hectare per capita on paper, prospered farmers have accumulated multiple hectares informally from others who succeeded less in irrigated paddy farming. Among 114 farmer informants, only 66, about 60 percent, were smallholders who held only one hectare. Middle-scale farmers with 2 and 3 hectares were 19 and 6 percent of the population, respectively, while 3 percent held lands larger than 3 hectares. The largest-scale farmer held more than 20 hectares and the second-largest 7 hectares; such few very large lands pulled the landholding average among informants up to about 2 hectares. Surveys provided further data showing that farmers with larger landholdings also had higher amounts of additional income, drew more on household-external labor, and owned more properties and livestock, pointing to class division among farmers. Income-generating assets were unequally distributed: all land-rich farmers with 4–5 hectares and more owned big farm machines (tractors and combine harvesters), which they rented out to other farmers for income. In addition, they owned additional businesses, such as the few restaurants, bars, guest houses in the village, and rice milling facilities, engaging in food trade. Middle-scale farmers with 2 and 3 hectares had smaller properties, such as power tillers and village huts, selling food and household goods, hired laborers to relieve household labor shortages, and only occasionally drew on cultivation loans. In comparison, all smallholder informants (with one hectare) took loans; they sometimes hired day laborers to complete difficult tasks during harvesting and sometimes worked as laborers for extra cash. Their property ownership was limited to a few hens and cattle, small maize plots in the drylands, and home vegetable gardens. These assets secured food and income to some degree but did not ensure farming investments when the harvest levels were low following drought periods. Smallholders depended on the harvest from season to season to subsist and reinvest in farming, and low harvest triggered land loss and redistribution to the wealthier classes.

To provide a systematic understanding of how harvest translated into income and potentially improved livelihoods in 2016–17, together with informants, I coined four income thresholds based on the average market prices (Table 1). That year, prices were lower than usual due to the grain export ban introduced by the government. This ban led wholesale paddy prices for SARO5 to halve from TSH1.1 million per ton ($484) to TSH0.55–0.6 million per ton ($242–264) in 2015. In 2016–17, prices slightly increased but remained low at TSH0.75 million per ton ($330). Based on this average price, the profit threshold, at which the income barely covered the production costs (TSH2.5 million), was 3 tons per hectare. Farmers defined a harvest at and below this level as crop failure because based on the early marketing season price of TSH75,000 per sack, 3 tons generated only TSH2.25 million per year and zero to minus profit after average production costs were withdrawn. This harvest used to be the subsistence threshold in 2014–15 based on higher prices (TSH1.1 million per ton) and used to create TSH0.8 million profit per year ($352) and TSH2,000 per day ($0.9 per day). However, it generated only $0.3 per day in 2016–17. Decreasing prices caused this harvest level to become insufficient for subsistence.

The new subsistence threshold, at which the profit made a (minor) contribution to subsistence, became 4 tons per hectare in 2016–17. This harvest generated only TSH1.25 million per year ($550) and TSH3,000 per day ($1.3) in the subsistence range, allowing farmers to access a limited range of food and cover only the basic livelihood needs. The stabilization threshold, which enabled farmers to make a sufficient profit to sustain their livelihoods and replenish their farm investment capacities for the next season (2017–18), was 6 tons per hectare. This harvest generated TSH2 million profit per year ($880) and TSH5,000 per day ($2.4), enabling farmers to afford the children’s education costs, access a broader range of food, undertake basic household renovations, and occasionally buy clothes and travel based on the local commodity prices. In addition, loan-dependent farmers could save some cash, thus needing to take smaller loans in the subsequent season.

The wealth threshold, only achieved through intensive farming (requires hybrid seeds, more than seven bags of fertilizers, and mechanized monocropping), was 8 tons per hectare. This is also the harvest level targeted by the government (i.e., 8–9 tons and above). A harvest in this range of 8–9 tons per hectare could indeed bring a profit of TSH3.5–4.25 million per year ($1,540–1,870) and TSH9.6–11.6 per day ($4.2–5.1), enabling farmers to generate farming investment for the subsequent season, overcome loan dependence, meet the household needs, save cash, and access a greater variety of food and some luxury goods. Smallholders sustaining this harvest level for a few years could potentially transform into large-scale farmers and diversify incomes into wealth-generating nonagricultural businesses rapidly, which have been SAGCOT’s vision. However, smallholders did not achieve these harvest levels.

Despite the shift to intensive farming, the mean harvest was 5.6 tons per hectare for the informants with harvest data (n = 81)—much below the 8–9 tons goal and not significantly above the pre-intensification range of 4–5 tons per hectare. The difference between the highest and lowest harvest was significant; the highest was above 9 tons per hectare and the lowest below 1 ton, significantly deviating from the mean. Based on the increased production costs (TSH2.5 million per hectare) and below-average market prices, farmers would have to harvest much above the usual levels to be able to sustain their livelihoods; harvesting the same amount of paddy as in the previous year put them into a lower-income group.

A comparison of the harvest and land data shows that smallholders harvested less paddy and generated less income than large-scale farmers (Table 2). All large-scale farmers harvested above the mean of 5.6 tons per hectare: 6–10 tons per hectare, which enabled them to maintain their high-income status. Only two farmers harvested above 9 tons. One of them was a large-scale farmer who used fertilizers intensively: nine bags above the standard of seven. The other held middle-scale lands, one of the few people that invested in land leveling and external labor to practice SRI, as confirmed by the local extension agent. Thus, the highest income group generating $4.2 and more per day were middle- and large-scale farmers only. Farmers with larger landholdings than smallholders invested more in intensive paddy farming and better coped with erratic weather. Middle-scale farmers fell into both high- and low-income groups. For example, one middle-scale farmer rented three hectares and cultivated middle-scale for the first time and fully on loans in 2016–17, aiming to fold his income over a single season despite hitherto being a smallholder. Harvesting as little as 4 tons per hectare of paddy on average due to the erratic weather, his debt multiplied by several hectares instead, leading him to consider renting out his plot and seek employment in 2018. This case shows that the confidence in the drought-tolerant seeds put some farmers at a disadvantage by encouraging them to neglect the weather risks.

Among smallholders with harvest data (n = 34), the main group of interest in this study, corresponding to 42 percent of farmer informants, no one harvested above 6 tons per hectare. The majority, 53 percent, harvested 4–6 tons per hectare and stayed in the lower-middle-income group ($0.6–2.4 per day). Though they survived erratic rains, a few of them struggled to subsist because of paying back loans with high interest rates. Because of submitting a large portion of their harvest to local financiers for loan payments, they also could not add value to crops by warehousing paddy (to sell at higher prices at the end of the season) or milling it to rice that increase income per unit of harvest as land-rich farmers usually do. The second-largest smallholder population with 38 percent harvested 6–7 tons per hectare at the stabilization range, joining the upper-middle-income group ($2.4–3.3 per day). They generated sufficient cash to repay their loans and invest in farming in the subsequent season, thus decreasing their loan dependence, which should be considered a livelihood improvement. Only to a limited extent, intensive farming succeeded in moving smallholders into the government-defined middle-income status of $3 per day. Because only the middle- and large-scale farmers harvested above 7 tons per hectare, earning $3.3 and more per day, intensive farming mainly benefitted the already better-off farmers.

A minority of the informant population (9 percent) had crop failures with 3 tons per hectare and less, putting them into the lowest income group ($0–0.6 per day). Farmers who harvested below the average argued that this was due to erratic weather. For example, the smallholder with the lowest harvest anticipated a harvesting delay to September 2017 (the usual harvest time is July) and collected only 3 tons of paddy with zero profit (Farmer 3, July 09, 2017). She held low precipitation responsible for the crop failure: when the rains ceased and most people harvested in July, her crops were still not ripe and partially died. Another smallholder who collected a low harvest of 3.6 tons also pointed to insufficient rainfall, which he had anticipated before cultivation started (Farmer 4, July 14, 2017).

Narratives from all informants (N = 114) on livelihood transitions spanning a longer period since 1998 show that leaving farming to become laborers is impoverishment rather than adaptation. This is because all accounts explaining such transition involved poor harvest and debt, impelling people to abandon the scheme, liquidate their assets, farms, and livestock to pay for debt and look for jobs. For instance, an informant claimed “a life-ruining debt” had driven her family to give away their one-hectare plot in the scheme: she had harvested 4 tons of paddy and submitted all of this to a moneylender to pay her YARA fertilizer debt, and she and her husband started working as day laborers for the wealthier scheme farmers (Farmer 5, December 10, 2016). The village chairperson reported that the 2015–16 drought similarly affected many other livelihoods, leading hundreds of farmers to lease out their lands; “they did not migrate from Madibira but stayed and looked for work at the irrigation scheme” (Key informant 5, December 23, 2016).

The most common and available types of work were day and seasonal wage labor in the scheme, which could not compensate for farming in terms of income generation. Wages for day labor varied depending on the task (e.g., canal cleaning TSH2,000, harvesting TSH50,000). Seasonal wages covered multiple tasks, including land preparation, seedling transplantation, and harvesting, and ranged at TSH500–750,000 ($220–330) per hectare per season, below the subsistence threshold in 2016–17. Only farmers who additionally rented out their formal one-hectare landholding in the scheme at the average lease rate of TSH750,000 generated income equivalent to TSH1.25–1.5 million ($550–660) in the subsistence range. Despite such income established some livelihood security, paddy farming could bring much more cash, up to “multiple millions of shillings,” as a local authority claimed (Key informant 6, June 11, 2017). Hence, farmers strived toward this goal even in the sight of weather adversity.

Once farmers became laborers, they found financial recovery difficult. The transition to employment was irreversible for many of them, even when they kept their landholding on paper. For instance, one laborer stated that she hoped to take cultivation loans “some time” again, but for the near future, she wanted to continue engaging in seasonal employment and petty commodity trading, considering these as securer income generation options (Farmer 6, December 10, 2016). Nevertheless, these options did not allow them to generate sufficient income to return to farming. Another laborer, a former farmer, asserted “a constant risk of collapsing” in his family and that he “constantly sought ways to escape this (poverty) cycle” but failed (Farmer 7, December 09, 2016).