Shunsuke Matsumoto^{1}, Tatsushi Tsuboi^{2}, Godfrey Asea^{3}, Atsushi Maruyama^{1}, Masao Kikuchi^{1} and Michiko Takagaki^{1*}  
^{1}Graduate School of Horticulture, Chiba University, Chiba, Japan  
^{2}JICA Expert, JICA Uganda Office, Kampala, Uganda  
^{3}National Crops Resources Research Institute, Kampala, Uganda  
Corresponding Author :  Michiko Takagaki Chiba University, Chiba, Japan Tel: +81473088934 Fax: +81473088934 Email: [email protected] 

Received January 09, 2014; Accepted February 10, 2014; Published February 13, 2014  
Citation: Matsumoto S, Tsuboi T, Asea G, Maruyama A, Kikuchi M, et al. (2014) Water Response of Upland Rice Varieties Adopted in SubSaharan Africa: A Water Application Experiment. J Rice Res 2:121. doi: 10.4172/jrr.1000121  
Copyright: © 2014 Matsumoto S, et al. This is an openaccess 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.  
Related article at Pubmed Scholar Google 
Visit for more related articles at Rice Research: Open Access
Whether a rice Green Revolution in subSaharan Africa becomes a reality critically hinges on how far productive upland rice cultivation diffuses in the region. In order to quantify the drought tolerance, the rate of water response and the contribution of yield components to changes in yield due to water availability of upland rice varieties used in subSaharan Africa, we conducted water application experiments in Namulonge, Uganda, using NERICA 4, NERICA 10, NARIC 2 and Yumenohatamochi, with five different levels of water application. We found that the NERICA varieties were most drought tolerant, followed by NARIC 2. Yumenohatamochi did not withstand the lowest amount of water application of 378 mm. The results suggested that the minimum water requirement was around 311400 mm per season for the three varieties used widely in East Africa, and around 420600 mm for Yumenohatamochi, an upland variety in Japan famous in its drought tolerance. It was estimated that an additional water application of 1 mm increased rice yield by 1112 kg /ha for the upland varieties tested. The high water response of upland rice was brought about by high water response of four yield components, among which the rate of grain filling contributed most to the increase in yield, followed by number of panicles/m2, number of grains per panicle and 1000 grain weight, in the order of the degree of contribution, for all the varieties tested.
Keywords  
Drought tolerance; NERICA; Minimum water requirement; Uganda; Yield component  
Introduction  
Since a series of NERICA varieties (interspecific Oryza sativa x O. glaberrima progenie) were developed in the late 1990s by the Africa Rice Center (then WARDA), upland rice has been expected to catalyze a rice green revolution in subSaharan Africawhere nearly 50% of land area planted to rice is upland, with the other half under rainfed lowland [1,2]. Such expectation has met mixed results due partly to the vulnerability of upland rice cultivation to drought [3,4].  
There have been studies on the impacts of water deficit, water stress and drought on the growth and yield of upland rice [513], but all of them are conducted in Asia and Australia using Oryza sativa cultivars. Studies on the relationship between water and the yield of NERICA varieties in subSaharan Africa have been burgeoning, but reliable studies on drought tolerance and water response are still scarce in spite of its importance in promoting upland rice cultivation in the region. The yield response of NERICA varieties under different levels of water availability is reported [14,15], but the water application levels in these studies are 7001200 mm per crop in the former and 17003000 mm per crop in the latter, both without controlling rainfalls.The relationship between yield and soil moisture using NERICA varieties is reported [16], but the soil moisture contents tested are 5070%. Farmers in sub Saharan Africa are attempting to plant upland rice in many upland areas that are prone to drought with rainfall less than 600 mm per crop season [2], and critical soil moisture contents for the growth of upland rice are 40% and below [8,13]. More studies that focus on lower levels of water availability are needed.A rice production manual for East African countries recommends upland rice cultivation using NERICA varieties in areas where 5day rainfalls of 20 mm for about 90 days (360 mm /crop) are available, based on the results of experiments that test the effects of water application on the yield of NERICA 4, without presenting any detail of the experiments [17].  
Information on the drought tolerance of NERICA and other upland rice varieties grown in subSaharan Africa and their response to water is critically important not only for rice research in the region, in particular for breeding droughttolerant upland rice varieties [18], but also for developing adequate cultural practices suited to the environmental conditions in the region and disseminating upland rice cultivation among upland farmers there. With a basic purpose of reinforcing related information for researchers, extension workers and policy makers in subSaharan Africa, this paper reports the results of waterapplication experiments conducted in Uganda, aiming at clarifying the drought tolerance and water response of upland rice varieties adopted in subSaharan Africa. More specifically, this paper intends to (1) elucidate drought tolerance in terms of the minimum water requirement of upland rice varieties adopted widely in sub Saharan Africa, (2) quantify how the upland rice varieties respond to changes in water availability, and (3) examine how yield components contribute to the increase in yield as water availability increases.  
Materials and Methods  
Experiments  
Experiments were conducted from March to July 2012 (126 days) in the National Crops Resources Research Institute (NaCRRI) in Namulonge, Uganda (latitude 00º30’46.4N, longitude 32º38’03.6E, altitude 1120 m), using wooden boxes placed in a glassroofed, screenwalled greenhouse that shut out rainfall perfectly but kept sunlight and atmosphere the same as outside.  
The planting boxes, measuring 1 m × 1 m × 0.7 m (width × length × depth), were made in such a way that the four walls were shielded by plastic sheet and the bottom was with holes the total area of which took one tenth of the bottom area, and filled up with soil taken from the top soil (050 cm deep from the soil surface) of an upland field in the NaCRRI. The soil composed of 55% of sand, 35% of clay and 10% silt, with pH of 5.6.  
Upland rice varieties used in the experiments were NERICA 4 (henceforth denoted N4), NERICA 10 (N10), NARIC 2 (Naric), and Yumenohatamoti (Yume).The characteristics of the varieties used are found inprevious studies: for N4 and N10, for Naric, and for Yume [1923]. N4, N10 and Naric are upland rice varieties formally released in Uganda [24], and hereafter grouped as ‘African varieties’ for the convenience of comparison with Yume, an upland variety in Japan famous in its drought tolerance.  
Seeds were sown by dibbling 6 hills at a space of 30 cm ×15 cm (22.2 hills /m^{2}) and 4 varieties were planted in a box.Fertilization was conducted following the recommended rates: NPK = 603030 kg / ha, 30 kg N, 30 kg P and 30 kg K applied as basal, 30 kg N applied at 30 days after sowing.Weeds were pulled out whenever emerged and no pest control was needed.  
Five treatments for water application were adopted: T1 = 21 mm /week (3 mm/day) x 18 (total water application = 378 mm), T2 = 28 mm /week (4 mm /day) x 18 (504 mm), T3 =35 mm /week (5 mm/day) x 18 (630 mm), T4 = 42 mm/week (6 mm /day)x 18 (756 mm), T5 = 49 mm /week (7 mm/day) x 18 (882 mm).The lowest water application treatment was set to be lower than the level of 4 mm /day, which is stated as the minimum rainfall for NERICA cultivation [17]. Actual water applications were made every Monday and Friday at the ratio of 4:3, for 18 weeks.Each water application treatment was replicated four times.  
Data were collected on the following parameters: (1) total aboveground dry matter weight at harvest, (2) yield, (3) yield components; number of panicles /hill, number of grains /panicle, rate of grain filling and 1,000grain weight, and (4) soil moisture content measured every day until 17th week after sowing by a soil moisture meter (DIK312A; Daiki Rika Kogyo Co. Ltd.). Yield componentswere fully enumerated for all the six hills, and yield and above ground dry matter weight were measured by uprooting the entire plants.  
Methods  
First, we examined the drought tolerance of the rice varieties by means of a simple correlation graph between yield and the total quantity of water applied, with fitted regression lines, distinguishing between ‘African varieties’ and Yume.Rough, but practically useful estimates of the minimum level of water required for the upland varieties were obtained by reading the coordinate in the graph of the intercepts made by the regression lines on the axis of the total quantity of water applied.  
Second, in order to quantify the response of yield, total dry matter and the four yield components to the total quantity of water applied, we estimated linear waterresponse functions for these factors. Taking yield as an example, the waterresponse function we estimate is expressed as follows:  
(1)  
WhereY = yield (kg/ha), W = total water applied (mm), V_{k} = dummy variable for variety (V_{k}= 1 if kth variety, = 0 otherwise), ε = random error, N = number of observations, and α,β,γ,δ are regression coefficients to be estimated. Note that Equation (1) has the same structure as the twofactorswithcrosseffect ANOVA model which can be expressed as , where μ = overall mean, ρ_{k} =effect of variety, φ_{w}=effect of water application, ω _{kw} =interaction effect between variety and water application, ε_{ikw}= random error, k = four varieties, and w = five water treatments.A difference in Equation (1) from the ANOVA is that water is treated as a continuous variable, not categorical as in the ANOVA.Indeed, the estimation of Equation (1) performs what the ANOVA model is designed to perform, and in addition, quantifies the effects of water, variety and their crossterm on yield [25]. A simple rearrangement of Equation (1) gives:  
(2)  
The second term in the right hand side of Equation (2) shows that the intercept of this water response function could be different by variety (intercept dummies), and the third term shows that the effects of water on yield (the slope of the response function) could be different by variety (slope dummies). In the estimation, all the explanatory variables were ‘centered’ by converting all the observations to mean deviations in order to avoid multicolinearlity [25]. Note that the overall intercept term (α) in a ‘centered’ regression is the mean of dependent variable, yield in this example. The estimation of Equation (1) was made for all the varieties and for ‘African varieties’ separately. In the estimation for all the varieties, dummy variables were set for N4, N10 and Naric, using Yume as the base of comparison.In the estimation for ‘African varieties’, dummy variables were set for N10 and Naric, using N4 as the base of comparison.  
Third, the contribution of changes in the yield components due to the changes in water availability to the change in yield is assessed by means of the additive decomposition of changes in a variable which is a product of other variables. Define yield as Y = Pâˆ™Sâˆ™Râˆ™G, where Y = yield, P = number of panicle/m^{2}, S = number of grains /panicle, R = rate of grain filling and G = 1000grain weight. Differentiating Y with respect to the total quantity of water applied (W) and dividing through the differentiated equation by Y, we obtain that is, the rate of change in Y is decomposed into the rates changes in P, S, R and G. Taking the derivative of Equation (2) with respect to W for P, S, R and G, and computing the rate of change for respective components for respective varieties, we can compute the relative contribution of these components to the changes in yield by variety.Since the decomposition equation is an approximation when differences (Δx) are used instead of differentials (dx), the lefthand side of the equation does not necessarily exactly tally with the righthand side.For the computation of the percentage contribution of the components, we used the summation of the rates of changes in the righthand side as the rate of change in yield. The decomposition was made for lower and higher water application levels separately, and for the ‘mean’ using the derivatives obtained from the estimated water response functions.  
As will be presented in the next section, Yume did not withstand low levels of water application, resulting in no yield in five replications. Since the inclusion of zeroyield observations results in overestimations of water response, we excluded these five observations, which made the total number of observations used in the analyses 75. Throughout the paper, we adopted three levels of significance levels for hypothesis testing, p<0.001, p<0.01 and p<0.05, with the symbols of ‡, † and *, respectively  
Results and Discussion  
The mean minimum daily temperature in the greenhouse during the period of the experiment was 15.7°C (standard deviation = 1.1°C), the mean maximum daily temperature was 28.4°C (1.3°C) and the average daily temperature was 22.1°C (0.9°C).  
Weekly average soil moisture contents were shown in Figure 1 for T1 (the lowest water application) and T5 (the highest water application), which ranged 6.9 – 13.8% with the mean of 9.9% for T1 and 10.1 – 20.0% with the mean of 14.5% for T5. The trends of the moisture contents at the later stages of growth indicated that the rice plants reached the harvesting stage by 14th week (98 days) after sowing. These levels of soil moisture contents are comparable to those of a study that uses upland rice varieties obtained from IRRI grown in chambers [8] and those of a study that uses NERICA varieties grown in open upland fields at IITA, Nigeria [15], but lower than the levels of soil moisture contents reported by previous studies conducted in Australia and Asia [7,13].  
The results of the experiment were shown in Table 1. At the lowest level of water application, N4 and N10 yielded 0.50 t /ha and 0.68 t /ha, respectively. Though very low, Naric also yielded 0.10 t /ha. However, Yume failed to reach the flowering stage. The yield of N4 and N10 jumped up to more than 3 t /ha at the second water application and that of Naric to 1.8 t /ha. In contrast, Yume gave no yield in one of replications at the second water application level, with the average yield in the rest of three replications of 0.35 t/ha, and it was at the fourth level of water application for the yield to exceed the 3 t /ha line. At higher levels of water application, yield reached around 6 t /ha for all the varieties tested. Total dry matter weight and all the four yield components also increased as the water application level went up.  
Drought tolerance  
Seventy five nonzero yield observations were plotted in Figure 2 against the total quantity of water applied, distinguishing ‘African’ varieties from Yume.Higher drought tolerance of ‘African varieties’ over Yume was apparent. The regression line for ‘African varieties’ intersected with the horizontal axis at 310 mm of water applied, and gave the yield of around 1 t /ha at the water availability of 400 mm per crop season. Taking into account the result that the rice plants reached maturity by 98 days after sowing (Figure 1), 311 mm (400 mm /126 days x 98 days) of water brought about the yield of around 1 t /ha for ‘African varieties’, which may be called the safe minimum water requirement. For Yume, the corresponding safe minimum water requirement was computed as 467 mm. These results suggested that ‘African varieties’ were more tolerant to drought than Yume by 33% in terms of the safe minimum water requirement. The results for ‘African varieties’ confirm that the recommended level of rainfall for NERICA cultivation of 360 mm per crop [17] is beyond the safe minimum level of 311 mm.  
We failed to find out any literature on the minimum water requirement of upland varieties used in subSaharan Africa to compare with our results. For outside Africa, it is reported that the threshold rainfall for upland rice cultivation in Asia and Latin America is 200 mm/month, or 600 mm for a crop season of three months [26,27]. Comparing the results of waterapplication experiments for upland rice conducted in Australia [28,29], USA [30] and Asia [9,13,31], it is shown that the minimum water supply level tested is 419 mm /crop [13]. Though for rainfed lowland ecosystem, it is reported that rice production is virtually impossible with rainfall below 450 mm/crop [32]. These observations point to the high drought tolerance of ‘African varieties’, of N4 and N10 in particular.  
It should be reminded that in our experiments water was applied regularly twice a week for 18 weeks.The distribution of rainfall during a crop season is not even as such.Since the water deficit at the reproductive stage affects yield more seriously than at the vegetative stage [57,12], a slight bias in rainfall distribution towards earlier or later stages could easily make the safe minimum water requirement of 311400 mm for ‘African varieties’ out of range.  
Water response  
The results of the estimation of water response functions were presented in Table 2. Water was a positive, significant factor for all the 12 regression equations estimated.Remarkable were high significance levels at which the coefficients of water were estimated: Except for Regression VIII for which the significance level was 5%, it far exceeded 0.1% for all the rest.  
First, let us look at the water response of yield for all the varieties (Regression I). The intercept of 3.46 t/ha showed the mean yield for the observations as a whole. The coefficient of water, 0.0123, the slope of the water response regression line for the base variety of Yume, indicated that a 1 mm increase in water availability increased yield by 12.3 kg /ha. The coefficients of three variety dummies for intercept were all positive and significant, indicating that the water response regression lines for these African varieties were located significantly above that of Yume, as shown in Figure 2 for ‘African varieties’ and Yume. The coefficients of two cross terms, Water x N4 and Water x N10, were negative and significant, indicating that the slopes of the water response regression lines for these NERICA varieties were less steep than that for Yume. These lesssteeper slopes for N4 and N10 were resulted from their higher drought tolerance that kept their yields high relative to Yume at lower levels of water application. Regression VII, the water response function for yield estimated by using only the observations for ‘African varieties’, showed few differences in their water response regression lines, except for the crossterm between Water and Naric. The slope of the regression line was steeper for Naric for the same reason mentioned for Yume.  
The estimated 1112 kg/ha/mm of water response of yield is high compared to earlier studies. Linear water response functions estimated for upland rice grown in North China show 5.2  5.7 kg /mm of water response of yield [10]. The comparison of six water application studies mentioned above gives the rate of yield increase of 2.5 kg/ha/mm [13]. A study using rainfed rice farmers’ field data reports that the rate is 1.2 kg /ha per 1 mm of total rainfall per season [33].  
Regressions II and VIII show that the rate of water response of total dry matter did not differ significantly among the varieties tested, though N4 and Naric produced significantly more total dry matter than N10 and Yume. Among African varieties N10 produced total dry matter significantly less than others. Although total dry matter increased significantly as water availability increased (Table 2), its rate of increase was far less than that of yield (Table 1), implying that the harvest index increased at the rate closer to that of yield.  
Number of panicle/m^{2} and number of grains per panicle were significantly larger for ‘African varieties’ than Yume (Regressions III and IV), while opposite was the case for 1000 grain weight (Regression VI). In contrast, there was no difference between ‘African varieties’ and Yume for the rate of grain filling, except for N10 that showed a significantly higher rate (Regression V). Also except for N10 for number of panicle /m^{2} and number of grains per panicle, the rate of water response (the slope) was not significantly different between ‘African varieties’ and Yume. For ‘African varieties’, the water response functions of four yield components differed little among the varieties. There were two exceptions for this; the rate of grain filling was significantly higher for N10 (Regression XI) and 1000grain weight was significantly heavier for Naric (Regression XII).  
Contribution of yield components to yield increase  
Yield increased significantly as water availability increased (Table 1). For ‘African’ varieties, the rate of yield increase was very high at lower levels of water application from T1 to T2 for N4 and N10 and from T2 to T3 for Naric, whereas for Yume, high rates of yield increase occurred at higher levels of water application from T3 to T5.The results of the decomposition of the yield increase were presented in Table 3. The decomposition based on the estimated water response functions in Table 2 (Mean) showed that the component that contributed most to the increase in yield was the rate of grain filling for all the varieties tested, followed by number of panicle/m2, number of grains per panicle, and 1000grain weight, in the order of the importance. The importance of the rate of grain filling was particularly large in the lower levels of water application. At higher levels of water availability, the contribution of number of panicles /m2 remained to be an important contributing component and number of grains per panicle increased its importance.  
These results are consistent with findings of earlier studies in Australia and Asia. It is found in Australia that water stress during panicle development reduces the rate of grain filling to zero [7]. It is found in Asia that the rate of grain filling is the key factor contributing to high harvest index of upland rice varieties under upland conditions [10], that decreasing water supply during 2040 days before heading reduces the number of grains per unit area and harvest index [12], and that the number of grains per panicle is the most important factor responsible for yield gap between aerobic and flooded rice [34].  
It is stated that a high potential yield and harvest index, as well as yield stability under different water regimes, are important putative plant characters for developing new elite upland rice varieties [12]. In addition to this, our findings suggest it would be necessary that two different, yet closely related, breeding strategies have to be sought in developing new upland rice varieties suited to subSaharan Africa. To enhance drought tolerance for wider dissemination of upland rice cultivation in the region, it is critical to build in a higher ability for grains to reach maturity. To enhance the yield potential of upland rice varieties to be planted in areas with relatively favorable rainfall conditions, it is effective to build in a higher ability for forming larger sink size (number of panicles/m^{2} and number of grains per panicle).  
Conclusions  
In view of the importance that upland rice varieties play in the rice Green Revolution in subSaharan Africa where many farmers are trying to plant rice in upland areas prone to drought, we conducted experiments in Namulonge, Uganda, to clarify water response of upland rice varieties using NERICA 4, NERICA 10, NARIC2 and Yumenohatamochi with five different levels of water application. We found that the NERICA varieties were most drought tolerant, followed by NARIC2. Yumenohatamochi did not withstand the lowest amount of water application of 378 mm that was applied evenly throughout the growing stages until harvesting. The results suggest that the minimum water requirement was 311400 mm per season for three varieties used widely in East Africa, and about 420600 mm for Yumenohatamochi. It was estimated that an additional water application of 1 mm increased rice yield by 1112 kg /ha for the upland varieties tested. The high water response of upland rice was brought about by high water response of four yield components. Among the components, the contribution by the rate of grain filling was highest, followed by number of panicles /m^{2}, number of grains per panicle and 1000grain weight, in the order of the degree of contribution, for all the varieties tested.  
Acknowledgements  
The authors are grateful to Dr. Kunihiro Tokida and the two anonymous reviewers for their helpful comments made on our earlier manuscripts.  
References  

Table 1  Table 2  Table 3 
Figure 1  Figure 2 