RESEARCH AT CIAT ON CASSAVA RESPONSES TO EXTENDED WATER SHORT-AGES

Responses to prolonged mid-season water stress: Research at CIAT (Connor et al., 1981; Porto, 1983; El-Sharkawy and Cock, 1987b) has shown that cassava tolerates a relatively long period of drought once the crop is established. In these studies, using a limited number of varieties, a stress period of 2-3 months was imposed 3-4 months after planting. The crop was later allowed to recover from stress for the rest of the growing cycle with the aid of rainfall and supplementary irrigation.

Figure 1 presents the relationship of yield with the seasonal average leaf area index (LAI) for four contrasting varieties, as affected by a prolonged mid-season water stress conducted in two separate field trials (El-Sharkawy and Cock, 1987b). The highest reference (100%) yields were 19 t ha^-1 dry root for CM

507-37 under nonstressed conditions over 345 d in Experiment I (El-Sharkawy and Cock, 1987b) and 11.2 t ha^-1 for M Col 22 under non-stressed conditions for 306 d in Experiment II (Connor et al., 1981). Compared with the control, the final root yields of the stressed crop were increased in vegetative vigorous varieties (e.g., M Mex 59); were reduced in less vigorous types (e.g., M Col 22); and slightly reduced or remained unchanged in varieties with intermediate vigor (e.g., the parent M Col 1684 and the its hybrid CM 507-37). These responses were closely related to leaf area characteristics (i.e., peak LAI and leaf area duration over the growing cycle), and to patterns of photoassimilate partitioning between top growth (stems and leaves) and storage roots. The vigorous types responded positively to stress because top growth was reduced and the harvest index was increased, whereas the less vigorous types responded negatively because leaf area was drastically reduced to levels below optimal LAI for root yield (El-Sharkawy and Cock, 1987b). The stability of root yields in intermediate types under both favorable and stressful environments stemmed from an ability to maintain leaf area near optimum during a major part of the growing cycle. Also, the relationship between leaf area and storage root yield is very important when cassava is subjected to a cold period coupled with water shortages in the subtropics where leaf area is reduced. Under these conditions, the crop requires a second warm-wet cycle for leaf area recovery and for attaining higher yields (Sagrilo et al., 2006). A similar ideotype approach was followed in studying tolerance to water stress in various crops including winter and spring wheat, maize, sorghum, millets, cowpea and coffee (Kirkham, 1980, 1988; Kirkham et al., 1984; Blum and Sullivan, 1986; Blum and Pnuel, 1990; Bolaños and Edmeades, 1993a,b; Bolanños et al., 1993; Whan et al., 1993; Richards, 2000; Hall, 2004; DaMatta and Ramalho, 2006). This information on the mode of response to water shortages is fundamental for a cassava breeding strategy and points to the need for selecting different plant types for different environments, a strategy later adopted by CIAT and IITA, (Hershey and Jennings, 1992; El-Sharkawy, 1993; Iglesias et al., 1995; Iglesias and Brekelbaum, 1996) and by national cassava programs, as in Brazil (Fukuda et al., 1992-1993).

Nevertheless, CIAT researchers needed to know to what extent cassava can tolerate a more prolonged period of water stress imposed at an earlier stage of growth. They also needed to simulate, as closely as possible, the common cassava-farming practice of planting cassava near the end of a rainy season, letting it pass through a long period of no rain, and then allowing it to recover in a second wet cycle. This objective was addressed using larger group of varieties.

Responses to prolonged early water stress: In the 1987-1988 season, eight cassava varieties (Table 1) were planted in a field drainage lysimeter (the total area, including borders, was about 3000 m²) at the CIAT research station at Santander de Quilichao, Cauca Department, Colombia, on 25 November 1987. Planting density was 12 500 plants ha^-1 in ridge, plot size was 25 m² and there were four replications per variety and treatment. Plants were adequately fertilized and the plots were kept weed-free manually. Because of rainfall deficits in December 1987 (94 mm) and January 1988 (81 mm), three irrigations were applied to ensure cassava sprouting and establishment. Two months after planting, before imposing the stress, supplementary irrigation was applied to bring the soil-water content to near field capacity within the 2.3-m soil depth. The available soil water within the 2.3-m profile of the experimental site was about 250 mm (between -0.03 and -1.5 MPa). Half of the experimental area was covered with white plastic sheets to exclude rainfall from day 60 to day 180 after planting. At this stage of growth, cassava had less than 0.8 LAI and less than 2 t ha^-1 total dry biomass, with no visible storage roots (Connor et al., 1981; Porto, 1983; CIAT, 1987-1989; El-Sharkawy and Cock, 1987b; Pellet and El-Sharkawy, 1993, 1997). During the stress period of four months, the control plot received about 540 mm of rain, together with three heavy irrigations within the first and second month, to compensate for the rainfall deficits in that period.

The total amount of water received by the control plot in four months was greater than the potential evapotranspiration at the Quilichao Experiment Station (about 4.2 mm d^-1). In the stressed plots, water was removed manually immediately after rainfalls and any cracks in the plastic sheets were sealed. The plastic cover was removed during the first week of June 1988. By the end of the stress period, the total water extracted from 2.3 m soil depth ranged among varieties from 168 to 200 mm. From 1 June to 20 October, the total rainfall was 656 mm. Supplementary irrigation to both the stressed and control plots was applied twice in June, July, and August to compensate for the low rainfall during that period. The total rainfall received from planting to harvest was 1406 mm in the control and 865 mm in the stressed plots.

During the stress period, field measurements of leaf gas exchange, both CO₂ uptake and H₂O loss, were made with a portable, infrared CO₂ analyzer (LI-COR model LI-6000) on single, attached, upper canopy leaves at solar radiation greater than 1000 µmol m^-2 s^-1 of PAR. These measurements were normally made between 0900 and 1300 h once a week within the first two months, and once every two weeks within the last two months of stress. Light interception, leaf water potential, and soil water content were also monitored. Final harvest was made on 20 October 1988 (11 months after planting) and the total standing biomass, root yield and dry matter contents of roots were determined.

Table 1 summarizes data of final standing total biomass, root yields, harvest index, and dry matter contents of storage roots. In the control plot, the total dried biomass ranged (in round figures) among varieties from 23 to 38 t ha^-1 , fresh roots from 42 to 60 t ha^-1, dried roots from 14 to 21 t ha^-1; dry matter contents from 32% to 39%; and the harvest index from 45% to 71%. In the stressed plot, the ranges were (again in round figures) 18 to 28 t ha^-1 for total biomass, 41 to 59 t ha^-1 for fresh roots, 14 to 19 t ha^-1 for dried roots, 30% to 38% for dry matter contents, and 62% to 76% for harvest index.

There were notable varietal differences in response to stress. Fresh and dried-root yields were decreased by stress in the parent M Col 1684, whereas they increased in its hybrid CM 507-37. Previous studies with these two genetically related varieties (CM 507-37 is a hybrid between M Col 1684 and M Col 1438) have shown that CM 507-37 is more vegetative vigorous and leafy (El-Sharkawy and Cock, 1987b, El-Sharkawy et al.,1992b). In other varieties, the yields were relatively unchanged, except for CM 2136-2, where dried-root yields decreased notably, mainly because of reduction in dry-matter contents from 35.5% to 31.1%.

Compared with the control, water stress across all varieties caused a reduction in total biomass by 12%, no change in fresh root yields, a reduction in dried-root yields by 3.4%, a reduction in dry matter contents by 3.3%, and an increase in the harvest index by 10%. These data clearly demonstrate cassava’s ability to tolerate prolonged water stress when it is induced gradually at an early stage of growth. Furthermore, the crop is able to recover and compensate, in terms of economic yields, from the adverse effects of stress. In areas with intermittent rainfall and with long periods of drought, cassava should produce reasonably well, providing good crop management (e.g., weed control and adequate fertilization) is practiced and cassava is grown in deep soils with good water-holding capacity.

The physiological mechanisms that underlie cassava’s tolerance of severe water stress are illustrated by data in Tables 2 and 3 and Figures 2, 3 and 4. The capacity of cassava leaves to fix atmospheric CO₂, a basic process for dry matter accumulation, during the stress period was 80% of that in the control (Table 2, Figure 2). This indicated that the photosynthetic process in cassava is not greatly inhibited by prolonged stress, an advantage that many other field crops do not have.

A second and important physiological mechanism of cassava leaves is their ability to partly close their stomata in response to water stress. For example, in the experiment, there was an average 43% reduction in leaf conductance to water vapor in stressed plots (Table 2) and consistent reduction over the stress period, compared to the control (Figure 3). The partial closure of stomata enabled cassava leaves to maintain, to some extent, the midday leaf water potential at levels comparable with those of cassava leaves in the control plot. Leaf water potentials at 1400 h across all varieties were -1.13 and -1.12 MPa for the control and stressed leaves, respectively (Table 3). This is another comparative advantage for cassava in dry areas, compared to other crops with poorer stomatal control.

In addition to its beneficial effect by preventing severe leaf dehydration, and consequently preventing impairment to photosynthetic capacity of the leaf, the partial closure of stomata reduces water loss through transpiration (Figure 4), thereby maximizing water-use efficiency, WUE (i.e., the amount of CO₂ fixed per amount of water transpired). Across all varieties, there was 39% increase in intrinsic leaf water use efficiency (P_N /g_s) in stressed crops over the control (Table 2).

A third and equally important physiological mechanism that enables cassava to withstand severe water stress is its ability to maintain a predawn leaf water potential comparable with that of unstressed cassava. In the experiment, leaf water potentials at 0600 h across all varieties were -0.39 and -0.40 MPa for the control and stressed plants, respectively (Table 3). This was partly achieved by reducing total leaf area (as indicated by the 31% reduction in light interception in the stressed plot, Table 2), thereby reducing total canopy transpiration, and by slow withdrawal of water from the deeper layers of the soil profile (Figure 5) (Connor et al., 1981; Porto, 1983; CIAT, 1987-1989; El-Sharkawy and Co
ck, 1987b; El-Sharkawy et al., 1992b; de Tafur et al., 1997a; Cadavid et al., 1998; El-Sharkawy, 2006a). During water stress, cassava fine roots extend for more than 2 m into deeper, wetter soil from where cassava can extract between 20% and 40% of its total water uptake (CIAT, 1987-1989; El-Sharkawy et al., 1992b). This is of paramount importance in areas with bimodal rainfall patterns and those with one short-wet annual period where excess water percolates deeper in soil profile and, hence, it could be extracted during long dry periods.

Also, it is possible that the phenomenon of “hydraulic lift” [i.e., nocturnal uptake of water from deeper wet soil layers that is transported and then released from fine roots into dryer top soil layers; see, for example, Mooney et al. (1980), Richards and Caldwell (1987), Caldwell and Richards (1989), Dawson (1993), Squeo et al. (1999)] occurs in cassava since predawn water potential in water-stressed plants always remained as high as in well-watered plants. This might be the case because the majority of fine roots exist in the top 0.40 m and a fewer portion of roots penetrate deeper layers (Connor et al., 1981; El-Sharkawy and Cock, 1987b) where a substantial water extraction occurs (Figure 5). Water uptake from upper layer (0.40 m) continued during long period of water stress with decreasing patterns over time (Figure 5; El-Sharkawy et al., 1992b, de Tafur et al., 1997a, El-Sharkawy, 2006a), thus indicating the existence of available water in this layer. Another characteristic that might be implicated in cassava tolerance to prolonged water stress is the obligate association with vesicular-arbuscular mycorrhizal fungi (VAM) (Howeler and Sieverding, 1983; Sieverding and Howeler, 1985). Among 20 cassava cultivars growing in large pots outdoors, the percent infected root length under stress varied from ~50 to ~90%, and these values were highly correlated with total plant root length across cultivars (r = 0.955, P < 0.001) (Sieverding et al., 1985). There is some evidence that plant-VAM associations may confer tolerance to water stress, particularly in species with low fine root density (Hayman, 1980; Nelsen and Safir, 1982; Ellis et al., 1985; Safir, 1985; Augé et al., 1987; Khalvati et al., 2005). Compared to cereal crops, cassava has a sparse fine root system and the extensive fungal hyphae-network in the soil may increase water absorption capacity of infected roots. By using a new split-root hyphae system in barley plants under well-watered and water-stressed conditions in growth chamber, Khalvati et al. (2005) found that water was transported from the fungi hyphae to barley root compartment under water stress, as compared to non-hyphae treatment. VAM-infected plants suffered less (relative to non-infected plants) from water stress in terms of leaf elongation rate, leaf turgor pressure, stomatal conductance and photosynthetic rate. These parameters indicate a better plant water status in VAM-infected plants.

A fourth mechanism underlying tolerance to drought is the ability of cassava to compensate partly for previous losses in dry matter production, due to water stress, by an increase in leaf canopy area (El-Sharkawy and Cock, 1987b; El-Sharkawy et al., 1992b) and by higher photosynthetic rates in the newly developed leaves after recovery, as compared to the unstressed plants (Figure 6) (El-Sharkawy, 1993, Cayón et al., 1997; de Tafur et al., 1997a; El-Sharkawy, 2006a). These higher photosynthetic rates in new leaves of previously stressed cassava were associated with higher leaf conductances, higher nutrient contents, as well as with stronger sinks for carbohydrate in storage roots (Cayón et al., 1997).

Not only can cassava tolerate long periods of soil water deficits aided with the above-mentioned inherent mechanisms, but it can also react to changes in atmospheric humidity (Figure 7) (Connor and Palta, 1981; El-Sharkawy and Cock, 1984, 1986, 1990; El-Sharkawy et al., 1984, 1985, 1989; Cock et al., 1985; Berg et al., 1986; El-Sharkawy, 1990, 1993, 2004, 2006a; Oguntunde, 2005, Oguntunde and Alatise, 2007). Cassava leaf stomata are sensitive to air humidity, irrespective of soil water content; they close rapidly in dry air when evaporation is high under field conditions, which may be translated into high leaf water potential. This mechanism enables cassava to maximize its WUE during periods of prolonged drought. When air humidity is high (e.g., early in the morning and during rainy periods), the stomata remain open. Thus, in a humid atmosphere and in the presence of soil water deficits, cassava leaves remain photosynthetically active and the crop can produce well; for example, in the Pacific coast of Ecuador, cassava produces 8 to 12 t ha^-1 of fresh roots with only 400 mm of rainfall in 3-4 months. In that region, the intensity of solar radiation is low because of cloudy skies and, hence, evaporation is low. A similar situation occurs in the Pacific coast of Peru where rainfall is very low but there is an intense fine mist that persists for hours, allowing stomata to remain partly open and the leaf to actively fix CO₂ at a lower rate of transpiration. Thus, WUE at the leaf level (CO₂ uptake per H₂O loss) and at the crop level for the whole growing cycle (dry matter produced per total water loss) are maximized in this case. Cock et al. (1985) found that increasing air humidity by fine misting from 1000 h to 1500 h, in a large cassava field experimental plot that was kept wet via irrigation and protected from wind drift by hedge rows of tall elephant grass at the CIAT Experiment Station, Palmira, Valle Department, Colombia, resulted in both higher leaf photosynthesis and higher root yields than in the adjacent unmisted plot that was equally irrigated. Moreover, leaf photosynthetic rate was significantly and positively correlated with air humidity, indicating stomatal reactions to air humidity even in a wet soil (Figure 8).

Coupled with stomatal sensitivity to air humidity is the strong leaf heliotropic response that allows leaves to track solar radiation early in the morning and late afternoon when the leaf-to-air water vapor deficit (VPD) is low. At midday when solar elevation is high and VPD is greatest, cassava leaves bend downward (i.e., leaf drooping movement) irrespective of soil water content and leaf turgor pressure (El-Sharkawy and Cock, 1984; Berg et al., 1986).
The net result of these two leaf movements is to maximize light interception and total canopy photosynthesis when WUE is greatest, and to minimize light interception when WUE is least.

In the present trial, the four months during which soil water stress was imposed coincided with a rainfall peak (total rainfall in April and May 1988 at Quilichao was 400 mm). During April and May 1988, the last two months of the stress period, the photosynthetic rates of the stressed plants were 60% to 70% of those in the control plants (Figure 2). This remarkable photosynthetic activity of the stressed cassava can be attributed partly to the favorable effects of high humidity which kept the stomata partly open (Figure 3). It may be concluded that cassava is extremely tolerant (or resistant) to prolonged drought because of multiple-inherent morphological, structural and physiological plant traits that allow the crop to obviate the negative effects of severe water stress.

The same trends in responses to extended water deficits imposed at early (2-6 months after planting), mid-season (4-8 months after planting) and terminal (6-12 months after planting) growth stages were observed in a 3-yr field trial with four contrasting cultivars that differed in their vigor (CIAT, 1992, 1993; Caicédo, 1993; El-Sharkawy et al., 1998; El-Sharkawy and Cadavid, 2002; El-Sharkawy, 2006a). Across cultivars there were no significant differences in root yield among water regimes, but there were significant differences among cultivars indicating genotypic x treatment interactions (P <0.01) (Table 4). Similar responses were observed in the Sudan Savanna zone of Nigeria using variation in the soil-water table as a variable for water availability (Okogbenin et al., 2003). These findings support the sound breeding strategy for developing cultivars for specific ecozones (Hershey and Jennings, 1992; El-Sharkawy, 1993; Iglesias et al., 1995; Jennings and Iglesias, 2002).

The discussed above plant traits and mechanisms that underlie cassava tolerance to prolonged drought have further implications for the possible expansion of adaptable cassava cultivars into marginal lands and under adverse climatic conditions. As a potential food and feed crop for the tropical and subtropical regions most likely affected by global climate changes, cassava will probably become an important food-security source in developing countries where there are severe food shortages (Rosenzweig and Parry, 1994; Kamukondiwa, 1996; El-Sharkawy, 2005; IPCC, 2006). Rosenzweig and Parry (1994) pointed out that cereal crop production in the tropics and subtropics will possibly decrease in the near future because of global climate changes, hence, food shortages would be further aggravated in these regions.

Selection for drought tolerance in cassava for seasonally dry and semiarid environments in Colombia

The physiological research, as discussed above, laid the foundation for improving the cassava genetic base, and for selection for drought tolerance in seasonally dry and semiarid environments where a significant portion of cassava production occurs (El-Sharkawy, 1993; Iglesias et al., 1995). A large group of cassava from the core germplasm collection was screened for leaf photosynthesis and productivity in seasonally dry and semiarid environments in Colombia.

Evaluation of core germplasm for productivity and photosynthesis in seasonally dry environmenst at the southwest Andean mountains of Colombia: In the 1986-1987 growing season, 127 CIAT cassava accessions, including cultivars, land races and breeding lines were screened on a private farm in the Patia Valle, Cauca Department, Colombia (600 m a.s.l., latitude 2º09’N, longitude 77º04’W, mean annual temperature 28ºC with little seasonal variation, average atmospheric humidity about 70%). The soils in Patia Valley are heavy clay, and the farm was under continuous pasture grasses, mainly Panicum maximum, for the last 25 years. The trial was planted at a population density of 15,625 plant ha^-1 on 23 October 1986, with adequate fertilization. The site received about 700 mm of rain in 309 d, but from December 1986 to April 1987, the rainfall was much less than the potential evaporation which was greater than 5 mm d^-1. The Patia Valley, lying between the central and western Andes mountains, is characterized by two wet periods (October-December and March-June), and with high solar radiation of about 22 MJ m^-2 d^-1. The 1986-1987 season was particularly dry, with no rainfall recorded from June to August. The trial was harvested on 26 to 27 August 1987.

Measurements of single-leaf gas exchanges were made with a portable infrared gas analyzer (LI-COR model LI-6000) using central lobes of upper canopy leaves on several occasions between February to June 1987. At this stage of crop growth, LAI was near its peak, and storage root bulking rate was greatest, and, hence, both carbon source capacity and root-sink demand were near optimal. Measurements were always made from 0900 to 1300 h when the solar irradiance exceeded 1000 µmol m^-2 s^-1 of PAR. Leaf water potential was determined with the standard pressure chamber technique (Scholander et al., 1965) on lobes from the same leaves used for gas exchange. Values for leaf water potential ranged from -1.0 to -1.5 MPa, across varieties and measurement dates. Nitrogen, phosphorus and potassium contents were also determined on the same measured leaves (El-Sharkawy et al., 1990).

Across all accessions, both fresh total plant biomass and dried root yield were highly significantly and positively correlated
with average leaf photosynthetic rate, and the correlation was higher in the high and medium top weight varieties than in the low top ones. These results indicate that, on the one hand, at high level of light interception (i.e., near optimum LAI in high and medium top weight), there was a direct relationship between productivity and leaf photosynthesis. On the other hand, at lower light interception because of lower than optimal LAI in the low top genotypes, the relation is weaker and light interception appears to be the predominant factor in determining productivity. Thus, when both canopy light interception and root-sink demand for carbohydrate are not limiting, productivity correlates well with leaf photosynthesis, as measured in the field (El-Sharkawy and Cock, 1990; El-Sharkawy et al., 1989, 1990).

Sixteen clones were selected from the many screened accessions on the basis of their high-yield performance, and were planted on 13 April 1988, in another adjacent private farm in the Patia Valley, and at a population density of 10,000 plants ha^-1. A split-block design with four replications was used to allow for two fertilization treatments, i.e., (1) without fertilization; (2) with 50, 100 and 100 kg NPK ha^-1. The size of the plot per clone was 25 m². The crop received about 950 mm of rain during the growing cycle of 308 d, with 560 mm out of the total rainfall occurring in October and November 1988, which resulted in a significant amount of water runoff. On 14-16 February 1989, the nine central plants per plot were harvested to determine biomass and root yield weight. Measurements of single-leaf gas exchanges were made only once per day on 29 August to 7 September 1988, with an LCA-2 portable infrared gas analyzer (Analytical Development Co., UK) that operates in an open-end system, in contrast to the LI-COR 6000 that operates in a closed fashion system. Across all blocks and fertilizer treatments, 35 fully expanded upper-canopy leaves were measured per cultivar. A small leaf chamber (Parkinson Broad Leaf Model) was clamped over the middle portion (6.25 cm² leaf surface area), in contrast to the 4-L leaf chamber used with the LI-COR closed system where the whole lobe of cassava leaf was measured for gas exchanges. All measurements were made from 0800 to 1300 h with solar irradiance exceeding 1000 µmol m^-2 s^-1 of PAR. Normal air with 320 ± 10 µmol mol^-1 CO₂ was drawn from above canopy using a vertically mounted 4-m glass-fiber probe connected to a pump.

Since no significant fertilizer effects were observed in root yield and gas exchange rates, data were pooled. Average root yields were higher than those in the 1986-1987 preliminary screening trial, and this was mainly attributed to the higher rainfall in 1988-1989, as well as to the smaller group of selected high-yielding clones. The mean dry root yield among the 16 cultivars ranged from 15 to 27 t ha^-1, indicating the high yield potential in cassava when grown in near optimal environments. In this trial with nearly 9000 m² land including borders, the overall average dry root yield harvested from the whole area exceeded 20 t ha^-1.

Evaluation of core germplasm in both seasonally dry and semiarid environments in northern coast of Colombia: Two field trials were conducted during the 1992-1993 season in two locations at the northern coast of Colombia using two groups of cassava clones selected from the CIAT core germplasm. One trial was conducted on a private farm at Santo Tomas, Atlantic Department (14 m a.s.l., latitude 10º57’N; longitude 74º47’W). At this site the mean annual rainfall of 830 mm is 50% of the mean annual pan evaporation of 1650 mm, with a rainy period from May to November and a dry period from December to April. The soil at the site is sandy (>80% sand) with low water holding capacity, very low in organic matter and low in nutrients. The second trial was conducted at a site of a religious school for the native young students (Aremasain) near Riohacha, Guajira Department (4 m a.s.l., latitude 11º32’N; longitude 72º56’W). At that site the mean annual rainfall of about 560 mm is 25% of the mean annual pan evaporation of 2300 mm. The rainfall distribution pattern in this region is characterized by a short rainy period from September to November, a dry period from December to April, and a second low-rainfall period from May to August. The soil in that site is sandy (> 80% sand) with low water holding capacity, very low in organic matter and nutrients. In both trials, no chemical fertilizer was applied.

Healthy stem cuttings (20 cm long) of the tested cultivars were planted on 25 September 1992 (Santo Tomas, Atlantic) and on 19 September 1992 (Riohacha, Guajira) on flat places of disked land at a 1 m by 1 m distance and in 5 x 5 m plots with four replications in a randomized complete-block design. The eight central plants from each plot were harvested for determination of yields on 4 August 1993 (Riohacha) and 5 August 1993 (Santo Tomas).

Under the above-mentioned stressful environments, average oven-dried root yield was 6.7 t ha^-1 at the seasonally dry location (yield ranged among cultivars from 5.8 to 7.6 t ha^-1), whereas at the semiarid location overall average yield was 2.3 t ha^-1 (yield ranged among cultivars from 0.4 to 3.3 t ha^-1). These levels of productivity, without fertilization and with severe prolonged drought, illustrate again the high adaptability of cassava to adverse atmospheric and edaphic conditions. Moreover, the crop not only survived but also produced reasonably well, where other major staple food crops like tropical cereals would not be able to compete with cassava. The most drought-tolerant tropical cereals such as grain sorghums and millets (Blum and Sullivan, 1986) perhaps would fail to produce under the semiarid conditions experienced in these trials. Nevertheless, because of the severe shortage of rainfall in the semiarid environment, root dry matter content was lower (less than 30%) than in seasonally dry environments. In practice, however, such as in northeastern Brazil with mean annual rainfall less than 700 mm, the crop is allowed to go into a second wet cycle that leads to higher yields as well as higher root dry matter content.

Measurements of leaf gas exchanges were made with an LCA-2 portable infrared gas analyzer during several days from February to March, 1993. All measurements were made on upper canopy leaves (four leaves per cultivar per replication with a total of 16 leaves) between 0800 h and 1200 h local time with a solar irradiance higher than 1000 µmol m^-2 s^-1. Measurements were taken during the dry period 4-5 months after planting at air temperatures within the leaf cuvette ranging from 29 to 37ºC, depending on time and date of measurements. This range of temperatures is near the optimum for photosynthesis in cassava as measured under controlled laboratory conditions (El-Sharkawy and Cock, 1990; El-Sharkawy et al., 1992a).

Breeding for drought tolerance in cassava under the semiarid conditions of northeastern Brazil: Besides being, for millennia, the main geographical site for the origin of cassava, the center for its genetic diversity and for its domestication (Allem, 2002), Brazil is the largest cassava producer in Latin America. According to FAOSTAT (1999) (cited by Henry and Hershey, 2002), the 1999 area harvested under cassava in Brazil was about 1.54 million ha, about 21% less than that in 1990 and the total fresh root production was 20.2 million tons (12.7 % of world 1999 estimated production of about 158 million tons). The estimate of root yield across the country was 13.1 t ha^-1 for 1999, slightly higher than that in 1990 (about 12.6 t ha^-1). On the one hand, one reason behind the decrease in acreage, and consequently the reduction in total production, was the pattern of decreasing cassava cultivation in the most favorable environments in southern Brazil (El-Sharkawy, 1993). On the other hand, in the more marginal regions of the semiarid northeastern Brazil, the area under cassava production expanded rapidly and now may account for more than 50% of the total Brazilian production. This trend strengthened the importance of improving the genetic base of cassava, and for breeding new cultivars more adapted to the severe water stress conditions prevailing in that region. This objective was further strengthened by the knowledge of cassava’s inherent potential for drought tolerance and the newly acquired basic physiological information and insights about the mechanisms underpinning such tolerance.

In late 1980 and early 1990, breeding efforts at CIAT were further integrated with the Brazilian national institutions involved in cassava research, mainly the federal research organizations of EMBRAPA and CNPMF, with headquarters at Cruz das Almas, Bahia State (Fukuda et al., 1992-1993). Also, collaboration with IITA, Nigeria, took place at the same time. As the crop physiologist at CIAT, I participated, along with CIAT breeders and the Brazilian national cassava team, in the initiation of a research project for cassava breeding in northeastern Brazil that was supported by the International Fund for Agricultural Development, Rome (El-Sharkawy, 1993). Based on the available meteorological data, four relevant sites were pre-selected for screening cassava germplasm in northeastern Brazil, namely: (1) Itaberaba, Bahia (270 m a.s.l., latitude 12º31’S). At this site, mean annual rainfall is about 718 mm, with a continuous rain throughout the year, but with two wet cycles. From January to April total rainfall is about 332 mm. There is a shorter wet cycle from November-December with a total rainfall of 200 mm. The rest of the year is considered dry as the monthly rainfall oscillated between 20 to 40 mm, which is far below the potential evaporation. (2) Quixadá, Ceará (179 m a.s.l., latitude 4º57’S). At this site, mean annual rainfall is about 677 mm, with only a four-month wet cycle (from February to May) with a total rainfall of 542 mm. The rest of the year is extremely dry, as monthly rainfall oscillated between zero to 45 mm. (3) Petrolina, Pernambuco (376 m a.s.l., latitude 9º22’S). At this site, mean annual rainfall is about 400 mm, with monthly distribution (mm) as follows: 50 in January, 78 in February, 92 in March, 43 in April, 7 in May, 4 in June, 2 each from July to September, 9 in October, 45 in November, 64 in December. This site is the driest among the pre-selected sites, as illustrated by the pattern of rainfall distribution. (4) Araripina, Pernambuco (816 m a.s.l., latitude 7º32’S). This site is the wettest among the pre-selected sites with mean annual rainfall of about 820 mm. The rainfall distribution pattern is very similar to that in Petrolina, but with the three wetter months having a total rainfall of 422 mm (January 114, February 134, March 174). The rest of the year was considered dry, as indicated by the monthly rainfall distribution that oscillated between zero and 63 mm. The soils in these sites are sandy with low water holding capacity in addition to being very low in fertility.