At a regional level, SSA experienced a 0.2-2.0°C temperature increase during the last several decades [2]. It is also predicted that a mean annual temperature in Africa would be 2°C higher during the middle of the 21^st century than the late 20^th century under medium scenarios. Despite uncertainty about rainfall changes in SSA, climate change models consistently predicted increasing incidences of drought [29]. As agricultural production in this region largely depends on precipitation and is temperature-sensitive, increases in temperature and decreases in rainfall likely affect rural livelihood as well as agricultural production in negative manners [2].

In Uganda, approximately 75% of farm households rely on rain-fed agriculture and few farmers employ an irrigation system [6, 7]. Negative impacts of drought have already manifested themselves on the country’s agricultural production. For instance, the droughts in 2006 and 2008-2011 affected food and livestock production that resulted in higher food prices. And the loss from the 2010-2011 droughts was estimated USD 1.2 billion [2]. In fact, a study found empirical evidence that below-average rainfalls and above-average temperatures in the long term may highly likely depress agricultural productivity in Uganda [7].

Maize is not exceptional to escape impacts from droughts and temperature increases. It was reported that droughts could cause up to 80% maize yield loss and a temperature-increase to 27°C decreased maize production [2, 28]. However, to take better measures to protect maize from the adverse weather shocks, it is important to gauge a degree to which the crop production is drought-vulnerable in the first place. A recent study attempted to estimate the drought vulnerability of maize production in Uganda under a predetermined set of precipitation and temperature [2]. In this study, vulnerability is defined as the degree to which a system is susceptible to and unable to cope and recover from adverse effects of extreme weather events and climate change. The concept of vulnerability is specified as a function of three to project future vulnerability of maize yield: maize sensitivity, maize exposure to droughts and temperature increase, and adaptive capacity of maize and farmers.

In detail, first, sensitivity is defined as a reduction in maize yield due to drought and temperature increase. Second, exposure is described as magnitude, intensity and duration of drought and temperature increase on maize yield. Third, adaptive capacity is defined as flexibility or ability of the maize production system to adjust to drought and temperature increase and to cope with their consequences. In the study, adaptive capacity is represented with two proxies, % of poverty and % of literacy rate as a material and human asset, respectively [2].

The overall result of the study indicated that the drought vulnerability of maize production was high or 0.6 (defined high if > 0.57) based on the 1960-2014 precipitation data. However, the drought vulnerability could be as high as 1.54 if the future temperature increased by 2.5°C. This level of vulnerability was over twice of 0.7 with a temperature change by 2.0°C. Among the three indices, the sensitivity defined as maize yield reduction noticeably increased from 1.16 to 2.0 with the temperature changes, whereas the exposure and adaptive capacity remained the same. At a geographical level, the vulnerability varied with latitudes of the maize production areas. The northern areas showed higher drought vulnerability compared to the southern areas. In the north, it ranged from 0.58 to 0.64 while in the south, the highest was 0.27. More importantly, the adaptive capacity decreased in the north while the vulnerability, exposure and sensitivity to droughts tended to increase in this area [2].

The study explains that this spatial disparity in drought vulnerability was probably due to multiple factors. First, a different precipitation pattern between the two regions may be an explanation: in the north of Uganda, precipitation is uni-modal (April-October) with lower levels of rainfall while bi-modal (March-May and September-November) in the south. In addition, the south has water sources to help enhance precipitation such as lakes Victoria, Albert and Edwards. Another important factor is a different socio-economic gap between the two regions. For instance, the poverty rate in the north was estimated 46.2%, considerably higher than 21.8% in the south. Generally, a high level of poverty limits access to key production inputs such as fertilizer, improved seed, and irrigation [2]. It may also limit access to financial services. Moreover, the literacy rate in the north ranged from 60% to 63% while markedly higher in the south from 63% to 75% with the national average 69.6% [2]. The lower literacy rate in the north may further limit access to key information on technical knowledge and skills to better cope with adverse weather events.

As a conclusion, this study suggests the adaptive capacity based on the two proxies (poverty and literacy) is the most important of the three indices to cope with droughts: the trajectory of climate change cannot be determined for the sensitivity and exposure whereas the adaptive capacity to climate shocks can be improved.

According to a survey [13], maize farmers in Uganda could be categorized into three depending on their farming characteristics: commercial, limited cash input and no cash input (Table 2). The survey found over 60% of the maize farmers utilized no improved input at all and the maize yield in this category was substantially low compared to the other two.

Given that most maize farmers are smallholders who are resource-constrained and subsistence-oriented, their decision to adopt improved maize such as DTM may well be influenced by perceived risks or uncertainties [22]. One of the most intuitive risks can be economic uncertainty. While farmer-saved local seed may cost none, improved seed including DTM can cost up to USD 3 per kg [12, 27]. When compensation of an additional cost to adopt DTM is uncertain, maize farmers unlikely consider adopting it and likely choose other options. However, a benefit-cost analysis with DTM in Uganda indicated DTM appeared economically viable compared to a local variety [27].

This benefit-cost study estimated net benefits, costs and their ratios to analyze DTM cultivation in economic terms. In doing so, five DTM varieties that are widely cultivated across Uganda (the longe 4, 5, 7H, 10H and MM3) and one local variety were tested. The result indicates that the average net benefit of DTM was 140% higher than the local (USD 820 for the DTM and USD 342 for the local). Among the DTM varieties, the longe5 and 7H brought the highest net returns or USD 1030 and 919, respectively, under the dry spell. These were approximately three times higher than the local with USD 342. For the benefit-cost ratio, the DTM was 2.9 on average, whereas that of the local was 1.75 [27]. This study found the DTM generated higher economic benefits under the trial conditions including ploughing both manually and by ox, seeding rate of 10 kg/acre, weeding twice and fertilizer application. Although the result from one study should not be generalized, the benefit-cost appeared favorable for DTM in Uganda, suggesting economic terms are not the sole reason for the current low DTM adoption.

From the farmers’ side, the DTM adoption process starts with them becoming aware of DTM existence as potential adopters [19, 20]. Then, the farmers who acquire information and understand the attributes of DTM make an adoption decision. When they decide to cultivate DTM as a trial, the seed must be physically available, accessible and affordable. With their DTM trial experiences, the farmers may continue or discontinue planting DTM. Finally, the DTM adopters may share information with other maize farmers and the cycle restarts [22, 23] (Fig. 1).

Fig. (1). Farmer adoption stages for improved seed adoption from [22, 23].

For the adoption process, some studies explored characteristics of DTM adopters and possible determinants in detail. The following section reviews these studies and their key findings are summarized in (Tables 3 and 4).

The first study analyzed 840 maize-growing households in 14 districts across the eastern, western, northern, and central regions of Uganda and observed differences in household characteristics between DTM adopters and non-adopters [23]. For an average household size, the adopters had a significantly larger household (6.9) than the non-adopters (6.27). Also, the adopters owned a larger land (2.2ha) and livestock herd (1.29) than the non-adopters (1.7ha for land and1.04 for livestock herd). Relatedly, the adopters dominated the top income category allowing 17% of them to save compared to 9% of the non-adopters. For the access to information, the adopters (53%) more likely had access to information than the non-adopters (37%), indicating that the adopters were more informed than the non-adopters. In terms of input use, the adopters more commonly used hired labor, manure, chemical fertilizer, and pesticide. For instance, 24% of the adopters applied fertilizer, while only 10% of the non-adopters applied it. In addition, the adopters applied a larger amount of fertilizer (14.01kg/ha) than the non-adopters (6.86kg/ha). The adopters also allocated a significantly larger plot to maize (0.64ha) than the non-adopters (0.47ha). The adopters reported a higher maize grain yield (1.7ton/ha) than the non-adopters (1.5ton/ha). However, the seed use rate was lower in the adopters (21.82kg/ha) than the non-adopters (26.71kg/ha). Overall, the sampled DTM adopters appeared to be better resource-endowed, had better access to information, cultivated more maize, applied more inputs, used less seed and harvested better yield.

Furthermore, the study identified possible determinants with significant coefficients by controlling other production factors in the sampled households. The adoption determinants identified included a larger household size, higher income, more information received on varieties and higher use of hired labor and manure [23].

A separate study with the same household data (yet 864 households) as the previous study was set to further predict potential DTM-adoption rates with specific conditions as well as to identify DTM adoption characteristics [22]. Although the results of the most socioeconomic characteristics were similar to the previous study as expected, some characteristics related to information exposure and seed accessibility were more detailed and informative. First, while the adopters were more exposed to information on new varieties, there was a difference in information sources: significantly more adopters (12%) received information from the government extension services compared to the non-adopters (2%). Yet, the major source of information for both groups (81% for the adopters and 76% for non-adopters) was a social membership such as cooperatives and farmer groups. Aside from the different characteristics in exposure to information, a large gap was observed in physical and economic access to seed between the two groups. The gap was 68.8% in physical access to seed and 86.8% in economic access to seed i.e. access to affordable seed. An additional observation that more adopters (12%) received free seed than the non-adopters (2%) probably contributed to the widened physical access gap to seed and the adopters’ better-endowment and wealth-related factors likely help them make trips and purchase DTM seed [22].

Albeit, in a smaller scale than the previous two studies, this study similarly identified key characteristics of DTM adopters, adoption constraints and determinants [28]. The study examined 190 households (95 DTM adopting households and 95 non-adopting ones) in Kamuli (south-east) and Masindi (mid-west) districts. An adopter for the study was defined as a farmer who grew DTM at least for two consecutive seasons and did not use recycled seed.

There were differences in socioeconomic characteristics between the two groups in age, income, education, land size, maize cultivation acreage and access to credit among others. In detail, the adopters tended to be older and more educated with higher income, better access to credit and larger farmland and maize cultivation area than the non-adopters. For input use, more adopters tended to use fertilizer (12%) than the non-adopters (1%), although the overall fertilizer use was limited in both groups. For awareness and information source, only 5.8% of the non-adopters were aware of the DTM and more of the adopters received relevant information from the government extension services and non-governmental organizations: 58% of the adopters received information from the two sources while 38% of the non-adopters did. In a further analysis with coefficients, key determinants of DTM adoption included a level of education, year of experience in maize farming, seed price, distance to seed source, awareness and frequency of drought occurrence [28]. The reason that drought occurrence frequency was included might be that 91% of the sample reported they experienced droughts at least twice in the last decade, thus could have suffered a reduced maize yield due to the droughts. And the sample households in these two districts identified a high seed price as a major barrier (57%) to DTM adoption followed by fertilizer requirement, lack of awareness and limited access to seed.

Overall, the three studies indicated that the DTM adopters appeared to be better endowed with production resources, be more informed, educated and experienced, have a larger household, use inputs more efficiently, earn higher income and harvest better yield. However, it was not empirically explained how these characteristics may relate to one another (for instance, causality) and what characteristics can weigh more for DTM adoption.

A large-scale study across six SSA countries (Ethiopia, Malawi, Tanzania, Uganda, Zambia, and Zimbabwe) examined determinants and constraints of DTM adoption [30]. The result averaged in the six countries showed that maize farmers were more likely to grow DTM if they highly valued grain yield, early maturity and drought tolerance. Similarly, the adopters were found to have more education, larger land, more exposure to information. The main barriers for DTM adoption included seed availability, information receipt, resource availability, seed price and seed attribute [30]. Those overall results drawn from the six countries appeared corresponding to the ones specifically identified in Uganda. It indicates other SSA countries might have similar DTM-adopter characteristics and adoption constraints.

The previous studies indicated that awareness of DTM seemed influenced by the household's ability to access and understand DTM information while seed accessibility might be essentially a function of wealth-related factors. Given that most maize growers in Uganda are smallholders, scaling-up DTM adoption partially depends on making seed widely available and affordable as well as relaxing the information bottleneck [22]. With this rationale, a study predicted potential DTM adoption rates under three conditions related to awareness, seed availability and accessibility: first condition-adoption conditional on exposure to DTM i.e. unrestricted awareness: second condition-adoption conditional on awareness and physically available seed i.e. unrestricted awareness-access: third condition-adoption conditional on awareness, physically available seed and affordable seed i.e. unrestricted awareness-access-affordability.

On the first condition, it was estimated that the DTM adoption in Uganda could have been up to 22% if the whole population had been aware of the DTM, instead of 14%, the sampled DTM adoption rate. On the second condition, the adoption rate could have increased to 30% if the seed had been physically accessible in addition to awareness. On the third condition, the adoption rate could have been up to 47% if the DTM seed was affordable in addition to awareness and availability. These predicted adoption gaps were 8%, 16% and 33% for each of the three conditions [22]. This suggests that it is feasible to scale up DTM adoption once a bottleneck in awareness, seed availability and seed affordability is relaxed. Yet, this result simultaneously implies that universal adoption of DTM in Uganda is highly unlikely even after the bottleneck is removed. This indicates that other obstacles need to be delivered for a wider DTM adoption.

One such obstacle might be competition with other non-DTMs with preferred attributes that are not currently incorporated in the DTM varieties [12]. In order to examine preferences of maize attributes, a study was conducted in Masindi and Iganga, the two main maize-producing areas in Uganda. Nine leading maize varieties were preference-tested, including four DTMs (longe 1, 4, 5, and 7H). Of the nine varieties, longe6H (non-DTM) was the most preferred (75%) on a hedonic scale followed by longe5 (69%). The key attributes for their popularity were relative early maturity, resistance to pest and disease and seed availability. As longe6H is non-DTM, its relative early maturity (125 days) may be attributed to the popularity with its ability to escape droughts. The importance of early maturity was also reflected in the farmer's willingness-to-pay: the farmers were willing to pay 56.5% more for early maturity. However, curiously, drought tolerance itself had no significant influence on their willingness-to-pay in this study [12]. Although longe5 has a shorter maturity (115 days) than longe6H, longe5 has a much lower yield (4-6 ton/ha) than longe6H (8 ton/ha). Additionally, longe5 is susceptible to northern corn leaf blight, while longe6H is resistant [12]. The overall result indicates that farmer preferences need to be considered in breeding DTM for further adoption, once again emphasizing a participatory breeding process.

In general, cultivating an improved variety is certainly more advantageous than an unimproved one. One of the advantages is the difference in their input elasticity, especially to fertilizer. Farmers in Uganda may well be aware of the positive relationship between fertilizer application and yield. Nevertheless, its application in maize production remains low with the nitrogen (N) as the main limiting component [31, 32]. A survey found the fertilizer application rate in maize in Uganda was far below the recommended rate (6 kg/ha for the surveyed application rate vs. 100 kg/ha for the recommended rate) and only 5% of the maize area received the recommended rate [13]. Low or lack of fertilizer application can prevent improved maize from fully expressing its potential. And the requirement of complementary fertilizer can discourage smallholders to adopt improved maize and it was indeed identified as a DTM adoption constraint [28].

Anecdotally, the low adoption of improved maize seed and fertilizer is partly due to low farmer confidence in their market quality. The government of Uganda in fact, recognizes the prevalence of counterfeit seed in the market across the country [25]. According to the Uganda National Seed Policy, 30-40% of the seed traded in the market was estimated counterfeit [6]. This is partly due to the unstable seed procurement system and weak public forces to regulate the counterfeit seed industry [25, 26]. Market quality of fertilizer is often questioned as well. Maize farmers as consumers do not observe their quality before using seed and fertilizer since both are experience goods. Therefore, maize farmer’s ability and confidence to know about the market quality of seed and fertilizer influence their purchasing decision [32]. Thus, a supply situation of the key inputs in quality can influence DTM adoption.

To empirically investigate this input quality issue in Uganda, a study examined the market quality of fertilizer and improved maize seed and assessed whether the quality of the two could explain the current low take-up of both [32]. The study randomly obtained 369 samples of urea fertilizer as market-purchased fertilizer (market fertilizer hereinafter) and 30 samples of predetermined hybrid seed as market-purchased seed (market seed). For comparison, authentic urea fertilizer and hybrid seed were obtained directly from the fertilizer wholesalers and seed companies, respectively. The study however, did not mention if the hybrid maize seed was DTM. Nevertheless, the result offers an informative view of the current market quality of both.

First, the surveyed farmers were indeed aware of the substandard quality of the market fertilizer. On average, they expected the market fertilizer to contain 38% less N compared to the authentic and only 1% of the farmers believed the quality of the market fertilizer would be the same as the authentic. Secondly, the farmers were aware that using fertilizer increase maize yield, only depending on the fertilizer source. They expected maize yield 1.47 ton/ha without fertilizer, 2.53 ton/ha with the market fertilizer and 5.23 ton/ha with the authentic. This perception of the maize farmers was in line with the following test results.

The test of the market fertilizer found that it contained 31% less N on average (31.8% N per kg) than the authentic (46% N per kg). Moreover, about 10% of the market samples contained even less than 50% N of the authentic, and only 1% was short of less than 10% N of the authentic. While there was a substantial variation in fertilizer quality across the market samples, their prices were homogenous. For seed quality, the average quality of the market hybrid seed was similar to the quality of a mixture of 50% of the farmer seed and 50% of the authentic hybrid seed.

In production terms, maize yield loss due to the low quality of the market fertilizer was found largest with the authentic hybrid seed (65 kg/ha) followed by the market hybrid seed (57 kg/ha) and farmer seed (49 kg/ha). In economic terms, 81.6% of the market fertilizer was found not profitable with the market hybrid seed. On the contrary, 100% of the authentic fertilizer with the authentic hybrid seed was profitable. In detail, the utilization of both market fertilizer and hybrid seed resulted in a negative mean return rate or -12.2%. In contrast, the utilization of both authentic fertilizer and hybrid seed resulted in 51% of the mean return rate. However, when the market fertilizer was applied to farmer seed, the mean rate of return was positive or 6.8% as farmer seed costs none or little. In comparison, that of the authentic fertilizer with farmer seed was 54.2% [32]. Together, these results imply the adoption of the two technologies currently available in the local market appears simply unprofitable and famers seem acting accordingly as a rational being in economic terms.

The result of this study offers three important observations [32]. First, the farmers were aware that the quality of the market fertilizer was inferior and that the fertilizer sources made differences in yield. Second, the prevalence of the low-quality inputs is highly likely in the current market of Uganda. Third, smallholder maize farmers may not adopt the improved seed and fertilizer due to negative or uncertain economic returns.

Suri (2009) also highlights that uncertainty in returns could be a high barrier to technology adoption: adoption may be low when returns are heterogeneous as well as low [33]. It is probable that the current maize input market of Uganda is characterized by an equilibrium of low-quality input, low-trust, and low-adoption [32]. Low quality of production input could come from multiple factors, including a weak regulation, adulteration, inappropriate handling, or storage. Circumstantial evidence suggests that adulteration by bulking fertilizer or dyeing seed to appear hybrid is not uncommon. Furthermore, the low and heterogeneous quality of the input is not correlated with its market price. This indicates that an important reference for market input quality is missing since a price is generally adjusted with quality [32]. This, in turn, highlights that the current market ambiguity needs to be cleared for effective DTM promotion.