Sensitivity of cassava (Manihot esculenta Crantz) clones to environmental changes

ELSEVIER

Field Crops Research 36 (1994) 213-220

Field Crops Research

Sensitivity of cassava (Manihot esculenta Crantz) clones to environmental changes Carlos A. Iglesias a'*, Fernando Calle a, Clair Hershey b, Gustavo Jaramillo a, Eloina Mesa a ~Cassava Program, CIAT, Apartado Aereo 6713, Cali, Colombia bP.O. Box62, Elm, PA 17521, USA (Received 24 February 1993; accepted 19 January 1994)

Abstract The performance of 15 cassava (Manihotesculenta Crantz) clones in 14 environments in Colombia was analyzed to determine the possibility of improving stability of root-yield in cassava in association with minimum acceptable yields. The specific objectives were to study the relationship among agronomic traits, to evaluate genotypic sensitivity to changes in the environment, and to characterize and determine the representativeness of the evaluation environments. Correlations found between root yield and related physiological or quality traits were in a favorable direction for breeding purposes. This indicates that when selecting for a complex set of traits, indices might be established with major emphasis on traits with high heritability and/or stability. For some traits (number of commercial roots and length of stem with attached leaves) the range of genetic variability was broader in favorable environments. Variation among evaluation sites was greater than variation across years. The results indicate that intermediate to low genotypic sensitivity in terms of cassava root yield and dry matter content can be combined with improved potential for dry matter production per unit area. In order to improve the performance and stability of cassava gene-pools, representative sites should be selected within the priority agro-ecosystem, to evaluate the genetic base for at least two years before selection is made. Association between the mean and the sensitivity coefficient for different traits was either nonsignificant or positive for breeding purposes. Improvement in the mean of traits can be made independently from, or in relation to, genotypic ability to react to environmental changes. Key words: Breeding; Cassava; Environmental variation; Genotypic sensitivity; Manihot; Stability

1. Introduction Cassava (Manihot esculenta Crantz) is generally grown by resource-poor farmers who are unable and/ or unwilling to assume large risks in relation to their crop (Cock, 1979). Stability of performance, particularly for root yield, is important for a farmer when adopting new cultivars. General stability o f perform*Corresponding author. 0378-4290/94/$07.00 © 1994 Elsevier Science B.V. All rights reserved

SSDIO378-4290(94)OOOO7-Y

ance can be measured by testing over contrasting locations, cultural practices, a n d / o r years or seasons at a given location. Cassava farmers usually plant in one location and have a reasonable control over cropping conditions. One o f the major risks faced by these farmers is year-to-year variation. Mean performance of a cultiva r over years often is less important to them than the minimum yield in an unfavorable year; consequently the importance of yield stability must be considered in association with minimum acceptable yields.

214

C.A. lglesias et al. /Field Crops Research 36 (1994) 213-220

For many subsistance farmers, cassava is a "security crop", producing some yield when other crops fail (Cock, 1983 ). In order to maintain that feature in new genetic material, cassava breeders need to adjust selection conditions and criteria so that they are representative of cultural practices for the majority of growers within a given ecosystem. Stability of performance usually depends upon the genetic ability of a variety to resist or tolerate various environmental-induced stresses (Hershey, 1984). Several methodologies have been used to evaluate genetic potential and stability in face of the environmental variation (Finlay and Wilkinson, 1963; Eberhart and Russell, 1966), and groups of lines that exhibit similar patterns of response across environments can be identified. Recommendations of varieties with specific or broad adaptation can be made for the range of environments under consideration. Sensitivity to environmental changes is a characteristic of a line that can be transmitted to the progenies (Blum, 1988). Selection for stability may involve a trade-off with selection for maximum yield potential (Rosielle and Hamblin, 1981). The relationships between cassava root yield and some physiological or quality component traits have been reported (Birader et al., 1978; E1-Sharkawy et al., 1992) from studies involving one location and/or one or two years. Some of the component traits were less affected by environmental variation than was rootyield. These traits may serve as indirect selection criteria to improve the yield potential and stability of cassava germplasm. Crop physiology has contributed to the assessment of adaptative traits and the selection for adaptation in several breeding programs (Shorter et al., 1991 ). Characterization of environments for testing breeding material is usually done on the basis of the average yield of tested lines. Cluster analysis is commonly used to define responses to environmental changes (Shorter et al., 1977). Classification of broad environments for cassava breeding has taken into consideration climatic (mainly rainfall and its distribution throughout the year) and soil factors (Hershey, 1984). Variation within those broadly defined environments can be best assessed by assigning different relative weights to traits of importance for the breeding program in order to select the most effective site for screening genetic variability for one or more traits.

The Cassava Breeding Program at CIAT has taken an integrated approach; to improve simultaneously dry matter yield potential, resistance to diseases and pests, soil and climatic constraints and root quality. In order to sample the variability within those broadly defined environments, several testing sites are used, increasing the cost of the breeding program. In this paper the stability of performance of 15 cassava clones over a range of environments in Colombia is analyzed with the following objectives: (1) to examine the correlations between root-yield and related traits; (2) to determine genetic differences in the sensitivity of root yield potential to changes in the environment, for the sample of clones being considered; and (3) to study the environments to determine the relative importance of temporal and spatial variations when selecting sites for cassava breeding.

2. Materials and methods The cassava clones used were seven hybrids and eight landraces (Table 1 ). Yield trials were grown at three locations in Colombia over five years (Table 2), using 25-plant plots in a randomized complete block design with two or three replications and evaluating the Table 1 Cassava hybrid clones and landraces used in this study Clone Hybrids CG 915-1 CG ll41-1 CM 3299-4 CM 3306-4 CM 3306-9 CM 3555-6 CM 3772-4 Landraces Bra 12 Bra 191 Col 22 Col 1468 Col 1505 Col 1684 Col 2215 Ven 77

Geneologyor name Bra 12×Col 1643 Mex 11 × Col 65 [ (Ven 305) x Ven 218] × Col 22 Col 22x (Col 655A× Col 1515) Col 22X (Col 655A X Col 1515) (Col 638 X Pan 70) × Col 22 [(Pan 70X Ext) x (Col 22 XCol 647)] x [(Col 22 XVen 270) ×Col 638] Unknown Amarela Casca Roxa Ayapelana Mantiqueira Verdecita CMC 163 Venezolana Unknown

215

C.A. Iglesias et al. /Field Crops Research 36 (1994) 213-220 Table 2 Characteristics of location and years for advanced cassava yield trials in Colombia Location

Palmira Media Luna Carmen de Bolivar

Rain (mm/season) 1986

1987

1988

1989

1990

Mean temp. (°C)

1170 1230

982 1332 1200

1151 1789 1466

945 1943 978

861 2130 1183

23.8 27.2 26.8

central 9 plants. Plant spacing at Palmira was 1 m by 1 m, for a plant density of 10000 plants/ha, while at Media Luna and Carmen de Bolivar it was 0.8 m by 1 m for a population of 12500 plants/ha. The trials at Palmira were planted in the second half of May and harvested at the beginning of the following March. In Media Luna and Carmen de Bolivar, trials were established during the first half of May and harvested at the beginning of February. No fertilizer was applied except at Media Luna where 50 kg/ha of each N, P and K were applied. A mixture of herbicides ( 1.5 to 2 kg/ha of Diuron plus 2.5 to 3 l/ha of Alachlor) was applied after planting, and it was complemented by weeding once or twice with hoes. The traits evaluated were: (a) at pre-harvest: total plant height (PH) ; length of stem with attached leaves (LS); and branching index measured as the height of first branching over the total plant height (BI); (b) at harvest: fresh root yield (RY); number of roots per plant, longer than 20 cm and heavier than 150 g (CR); harvest index measured as the ratio of root yield to total harvested biomass (HI) ; and ( c ) after harvest: root dry matter in percentage (DM) ; and cyanide in root parenchyma (HCN), determined by the picrate method (Williams and Edwards, 1980) and scored on a scale from 1 ( < 10 ppm) to 9 ( > 150 ppm). The RY data were transformed to square roots and LS to logl0 in order to achieve normality in the distribution of the data. Simple correlation among traits was analyzed. The modified joint regression method for incomplete data proposed by Digby (1979) was used to estimate differences in the sensitivity of cassava clones to environmental effects, employing the following model:

Altitude (m )

1000 10 152

Soil texture

Clay Sandy loam Clay

Soil composition % Organic matter

P ( ppm )

K ( meq / 100g)

Ca ( meq / 100g)

4. I 0.4 2.2

6.2 5.8 7.5

0.75 0.06 0.90

20 1 52

Yij = T~ + B, Oj + Eij; i = 1,2 ..... n genotypes and j = 1,2 ..... n environments; where: Yij = performance of genotype i in environment j, Ti= expected mean for genotype i, Bi = sensitivity coefficient, 0j = effect of environment j, and E~j= error term. The environmental effect was estimated through an iterative process using yield trial means, with the restriction of E0 = 0. The characterization of environments was done through principal component analysis based on the trial mean for all the evaluated traits. Grouping. of environments was done using Ward's method for cluster analysis (Anderberg, 1973).

3. Results and discussion

3.1. Relationship among traits The RY was significantly correlated with all other traits except BI (Table 3). The relationship with HI was not as high as expected from previous reports for cassava (Kawano, 1990). The low but significant correlation between RY and HI may reflect a confounding correlation as both traits have a common base (RY). The HI has been used as a primary selection criterion for several cycles by CIAT's Cassava Breeding Program (Kawano, 1990). The improved clones derived from the latest populations had a HI of 0.5 to 0.6. This is probably the optimum level because at higher values of HI, root production potential is affected as a result of reduced photosynthetic area (Cock and E1-Shar-

C.A. Iglesias et al. / Field Crops Research 36 (1994) 213-220

216

Table 3 Correlation among evaluated traits for 15 cassava clones in 14 environments

RY

DM

HCN

PH

BI

HI

CR

LS

0.45 a 0.0001 b

- 0.31 0.0001 - 0.50 0.0001

0.52 0.0001 0.07 0.2950 -0.37 0.0001

- 0.03 0.6411 0.10 0.1732 -0.13 0.0757 - 0.06 0.3698

0.18 0.0107 - 0.01 0.8831 0.12 0.0740 - 0.25 0.0003 -0.15 0.0436

0.54 0.0001 0.20 0.0040 -0.19 0.0068 0.40 0.0001 0.01 0.8647 0.43 0.0001

0.51 0.000 I 0.40 0.0001 -0.36 0.0001 0.41 0.0001 -0.01 0.9855 - 0.09 0.1843 0.16 0.0249

DM HCN PH BI HI RC

aPearson correlation coefficient. bProbabilty > I R I, under Ho: RH o.

kawy, 1988; El- Sharkawy et al., 1990). Supporting this is the high correlation between RY and both PH and LS which are indicators of foliar development and leaf retention, respectively. The LS has become a primary trait for selection within CIAT's program, particularly for seasonally dry agro-ecosystems. Those clones with the ability to withstand long periods of drought (3 to 6 months), maintaining a leaf area close to the optimum, will produce more and be less affected in terms of dry matter content, as they will be using fewer root reserves for regrowth when the rainy period starts (Connor and Cock, 1981 ). The LS is an important trait to be considered when selecting for yield potential and stability at the same time. The CR is a direct measure of sink strength for a given genotype (Wholey and Cock, 1974). Its positive relationship with RY indicates that both top growth and sink strength need to be improved simultaneously, maintaining the proportion between them (i.e., HI) at an optimum level (0.5 to 0.6). Recent studies (Pellet, 1992) using reciprocal grafting of clones widely different in root number confirmed that both source (leaf area and photosynthetic rate) and sink strength (root number and size) are significantly correlated with RY. The RY was positively correlated with HI, PH, LS, CR and DM, and negatively with HCN. This indicates that for an integrated cassava breeding program, which

takes into consideration a complex set of traits as selection criteria, indices could be established to exploit these associations and select for those traits with high heritability and/or stability.

3.2. Environmental sensitivity Mean performance and sensitivity coefficients for all traits evaluated for the 15 clones are presented in Table 4. A high mean performance, together with a sensitivity coeficient equal to or lower than one, is desirable for yield potential, as cassava is grown mainly in marginal environments. Fig. 1 presents the mean and sensitivity coefficients for each of the 15 clones in terms of RY. Clones CG 915-1, CM 3306-4 and BRA 12 can be considered as having specific adaptation to favorable environments, while clones CM 3372-4 and CM 35556 can be recommended for lower yield potential environments. Clones CG 1141-1, CM 3299-4, CM 33069 and BRA 191 have a high mean RY and average sensitivity to environmental changes. They produce fairly well under a range of environmental conditions and can be indicated as having broad adaptation in terms of RY potential. The association between sensitivity coefficients and RY potential was positive but not significant (r = 0.23), suggesting the possibility of improving both traits independently. A separation of RY means for clones at each location

217

C.A. Iglesias et al. /Field Crops Research 36 (1994) 213-220 Table 4 Mean and coefficient of sensitivity (B0 for eight traits in each of the cassava clones evaluated Clone

RY (t/ha)

DM (%)

HCN (ppm)

PH (cm)

LS (cm)

BI (cm/cm)

HI ( g / g )

CR (No.)

Mean B i

Mean

Bi

Mean

Mean

Bi

Mean

Bi

Mean

Bi

Mean B i

Mean

Bi

CG915-1 CG 1141-1 CM 3299-4 CM 3306-4 CM 3306-9 CM 3772-4 CM 3555-6 BRA 12 BRA 191 COL22 COL 1468 CO1 1505 COL 1684 COL 2215 VEN77

21.4 18.9 18.4 17.1 18.1 19.4 16.0 18.7 18.5 14.8 15.7 16.2 14.8 12.2 10.2

1.49 1.02 0.88 1.29 0.98 0.76 0.62 1.53 1.00 1.00 0.85 1.07 0.91 0.80 0.62

32.6 36.9 35.1 38.1 37.9 33.2 34.0 31.7 34.7 32.5 28.8 34.9 30.2 37.2 28.1

0.97 0.61 1.15 0.97 0.98 0.92 1.03 i.09 0.84 1.48 1.15 0.49 0.63 0.85 1.71

6.8 7.0 5.2 6.2 6.1 6.6 5.3 8.0 5.5 7.1 6.9 6.4 8.7 5.5 7.0

1.50 0.59 1.23 0.77 2.25 0.58 1.54 1.04 0.48 1.95 0.81 0.43 -0.01 0.60 0.81

190 161 211 196 172 201 191 197 208 140 210 203 161 163 229

1.15 0.70 1.21 0.96 0.82 0.74 1.11 1.08 0.64 0.95 1.13 1.09 1.27 0.83 1.23

20.8 19.0 29.6 24.2 21.2 23.0 25.7 24.0 32.6 19.4 22.6 20.8 16.8 24.8 19.2

1.03 0.68 1.68 1.23 0.97 1.00 1.04 1.05 1.46 0.81 0.90 0.95 0.42 0.92 0.71

1.19 1.21 1.20 1.16 1.23 1.23 1.24 1.24 1.21 1.21 1.20 1.28 1.12 1.22 1.24

0.89 1.56 1.08 -0.10 0.80 1.53 1.69 0.59 1.16 0.51 1.05 1.19 0.79 0.95 1.46

0.61 0.54 0.53 0.54 0.59 0.58 0.52 0.52 0.50 0.63 0.49 0.52 0.55 0.48 0.47

1.43 0.48 1.02 1.37 1.29 1.01 0.77 1.07 1.86 0.94 0.74 0.20 0.07 0.44 1.71

3.85 3.14 3.17 3.35 3.64 3.78 3.00 3.42 3.10 2.71 2.71 3.14 2.42 2.02 2.72

1.07 0.94 0.87 1.50 1.55 0.83 1.01 1.43 1.30 0.99 0.79 0.86 0.40 0.51 0.49

S.D.

2.9

0.27

3.1

0.31

1.0

0.61

24.3

0.20

4.2

0.31

0.04

0.05

0.52

0.50

0.36

Bi

8 e N I l l l v l l y coil flclent

1,4

m

I

l

1,2 rmf

=lmlA

1

nw OIA IQIIe-~

0,8

I

~N rr

0,4

B

ov ~ - e

0,6

i

I

i

I

I

10

11

12

13

14

I

I

I

1,5 16 17 RY ( t / h a )

I

I

I

I

I

18

10

20

21

22

23

Fig. 1. Sensitivity of root yield of 15 cassava clones to changes in the environment.

across different years (analysis not shown) identified the best clones for each location. Those clones were: CG 915-1, CM 3306-4 and BRA 12 for Palmira; CM 3306-4, COL 2215, CM 3306-9 and CG 1141-1 for Media Luna; and CM 3306-9, CM 3306-4 and CG 1141-1 for Carmen de Bolivar. Clones with broad adaptation like CM3306-4, and to a lesser extent CM 33069 and CG 1141-1, are the ones sought by most cassava breeding programs. Locations with higher production

0.47

potential, like Palmira, require the selection of other specifically adapted clones such as CG 915-1 and BRA 12. In the case of Media Luna, a local clone (COL 2215) showed specific adaptation in terms of RY, as was expected. The majority of clones with high DM showed average to low sensitivity to the environment ( r = 0.48, significant at 10% probability). Given the positive relationship between RY and DM, it may be possible to increase DM production potential and at the same time reduce the sensitivity of that potential to changes in the environment. The potential of a genotype to produce HCN is an important feature that will determine both its end use and potential health hazard in case of direct consumption. The ability to accumulate cyanogenic glucosides in the root parenchyma seems to be highly heritable, yet sensitive to environmental influence (Cooke et al., 1978). The most desirable combination for breeding purposes is low HCN and low sensitivity as is the case with clones BRA 191, COL 1505, COL 2215, CM 3306-4 and CM 3772-4 (Table 4). A high positive association (P < 0.01 ) between the mean and sensitivity coefficient for RC (r = 0.66) and LS ( r = 0 . 9 0 ) suggests that screening for these traits will be more effective under favorable environments,

218

C.A. Iglesias et al. / Field Crops Research 36 (1994) 213-220

where the expressed range of genetic variability would be the broadest. The expression of genetic variability in these traits is apparently limited under unfavorable conditions (Kawano et al., 1978), and despite the positive relationship with RY, attempts to use them as indirect selection criteria may not be useful in increasing cassava yield under unfavorable environmental conditions. Cassava plants must develop sufficiently in order to reach an optimum photosynthetic area to sustain economic root yield and at the same time provide enough planting material for the farmer to establish the following crop cycle. On the other hand an excessive vegetative response to improved environmental conditions may lead to lodging problems (Cock, 1976). If the objective is to increase top growth, it will have to be combined with low sensitivity to environmental variation, which seems possible, as there is a nonsignificant correlation ( r = 0.28) between the mean for PH and the sensitivity coefficient. Performance such as that of clones CM 3306-4, CM 3772-4 and BRA 191 is the most appropriate. For an efficient cassava plant architecture, increase in top growth has to be accompanied by intermediate to late branching (high BI), (Cock et al., 1979). An increase in the BI mean, together with an intermediate to high sensitivity to changes in the environment, is desirable in order to maintain some branching capacity under the most unfavorable environments. This may lead to a canopy structure with more efficient use of incident light, as well as enhanced competition with weeds (Cock et al., 1977). The positive association between the mean and coefficient of sensitivity for BI (r = 0.46; P < 0.10) may partially facilitate obtaining such an objective. Hybrid clones and the most elite germplasm accessions (BRA 191 and COL 1505) had optimum values for HI (0.5 to 0.6) (Kawano, 1990). Based on the previous discussion for RY, CR and PH, an intermediate sensitivity for HI should result in lower HI for unfavorable environments, allowing an adequate balance between RY and top growth, and high HI under favorable environments where the balance would favor RY. As there is no association between the mean and environmental sensitivity for HI (r = 0.01, nonsignificant), it is possible to maintain HI at an optimum level and with intermediate sensitivity.

3.3. Characterization of environments

Environments are usually grouped using general climatic and soil information. Description of environments using data generated from trials should have a better prediction value since it takes into consideration the environmental and the genetic background being evaluated. Such description can be done considering just one trait (i.e. root yield) or a combination of two or more characteristics. In this case, the 14 environments were characterized using principal component analysis, considering the trial mean for all the evaluated traits. The following three components explained 82% of the total variability among environments: Component 1 = 0.46RY + 0.44DM - 0.42HCN + 0.002HI + 0.31CR + 0 . 4 0 P H - 0.03BI + 0.38LS, Component 2 = - 0.02RY - 0.25DM + 0.04HCN + 0.69HI + 0.46CR + 0.12PH - 0.44BI - 0.20LS, Component 3 = - 0.07RY - 0.004DM - 0.21HCN + 0.29HI + 0.36CR - 0.27PH + 0.80BI - 0.10LS. Component I reflects the general relationship among traits ( see Table 3), with relatively low importance for HI and BI. Component 2 involves high positive values for HI and CR together with high negative values for BI; while component 3 is mainly characterized by high values of BI. Once the 14 environments were characterized by these three components a cluster analysis was conducted. Four groups can be clearly differentiated, explaining 75% of the variability among environments. Table 5 presents the environments that fell within each group and the mean for the evaluated traits in each. Group 4 is a high DM-yielding potential environment; the other three can be classified as low potential environments and can be differentiated by their CR, PH and BI. Three of the groups (2, 3 and 4) reflect differences among evaluation sites across years. According to this, spatial variation appeared to be more important than temporal variation. There is a clear difference between Palmira on the one hand and Media Luna and Carmen

C.A. Iglesias et al. /Field Crops Research 36 (1994) 213-220

219

Table 5 Groups of environments based on cluster analysis and mean for the evaluated traits in each group Environments

Mean forevalu~edtraits RY DM HCN PH LS BI HI CR

Group 1

Group 2

Group 3

Group 4

Site a

Year

Site

Year

Site

Year

Site

Year

PM PM ML CB CB

1988 1990 1987 1987 1989

CB CB CB

1986 1988 1990

ML ML ML ML

1986 1988 1989 1990

PM PM

1987 1989

13 33.5 6.97 168.8 21.9 0.51 0.49 2.05

14.41 32.91 6.54 181.1 20 0.57 0.58 3.63

15.75 32.52 6.76 201.3 15.5 0.35 0.58 3.38

31.78 38 5.57 230.1 45 0.48 0.55 4.23

aPM = Palmira; ML = Media Luna; CB = Carmen de Bolivar.

de Bolivar on the other hand. Palmira is at a higher altitude and has more fertile soil than the other two locations. The difference between Media Luna and Carmen de Bolivar is related to higher soil fertility in the latter location. Abnormal distribution of rainfall in some years may mask differences in soil fertility or altitude among locations. Cassava is classified as a crop tolerant to water deficits, but shortage of water in the early months of establishment can irreversibly affect plant establishment and development (Connor and Cock, 1981). Group 1 represents the group of the worst environments, meaning that when a growing season is extremely unfavorable for cassava, the differences among sites tend to disappear, as also seems to be the case for genotypic differences (Rosielle and Hamblin, 1981 ). In this case it would be important to consider a combination of sites and years when measuring genotypic sensitivity to environmental changes. This information seems to support the Program's current strategy for breeding cassava for stress environments. Elite clones are selected after at least two years of evaluation at the three indicated sites, making it possible to account for the variation among locations and, partially, for year-to-year variations. Based on the analysis of information from a sample of clones and environments, it may be concluded that cassava yield can be increased more effectively by

using a set of selection criteria that includes some empirical physiological and quality traits in addition to root yield potential. For some of these traits (i.e., CR and LS), the range of genetic variability may be better expressed in intermediate to favorable environments. Variation among evaluation sites is more important than temporal variation. Selection for intermediate to low genotypic sensitivity in terms of RY and DM appears to be a useful objective when breeding cassava for less favorable environments. This objective could be better achieved when testing is done under representative conditions (i.e., Media Luna and Carmen de Bolivar). Association between the mean and the sensitivity coefficient for different traits was not significant for the majority of the traits analyzed, suggesting that it is possible to improve the mean of those traits independently from the genotypic ability to react to environmental changes. Sensitivity analysis across a set of environments may contribute to the overall objective of improving cassava dry matter production potential under less favorable growing conditions, while maintaining the capacity of the crop to respond to favorable environments.

References Anderberg, M.R., 1973. Cluster Analysis for Applications. Academic Press, New York.

220

C.A. lglesias et al. / Field Crops Research 36 (1994) 213-220

Birader, R.S., Rajendran, P.G. and Hrishi, N., 1978. Genetic variability and correlation studies in cassava (Manihot esculenta Crantz). J. Root Crops, 4: 7-10. Blum, A., 1988. Plant Breeding for Stress Environments. CRC Press, Boca Raton, FL. Cock, J.H., 1976. Characteristics of high yielding cassava varieties, Exp. Agric., 12: 135-143. Cock, J.H., 1979. Cassava research. Field Crops Res., 2: 185-191. Cock, J.H., 1983. Cassava. In: W.H. Smith and S.J. Banta (Editors), Symposium on the Potential Productivity of Field Crops under Different Environments, 1980. IRRI, Los Bafios, Laguna, Philippines, pp. 341-359. Cock, J.H. and E1-Sharkawy, M., 1988. Physiological characteristics for cassava selection. Exp. Agric., 24: 443--448. Cock, J.H., Wholey, D.W. and Gutierrez, O., 1977. Effects of spacing on cassava (Manihot esculenta Crantz). Exp. Agric., 13: 289299. Cock, J.H., Franklin, D., Sandoval, G. and Juri, P., 1979. The ideal cassava plant for maximum yield. Crop Sci., 19:271-279. Connor, D.J. and Cock, J.H., 1981. Response of cassava to water shortage: II. Canopy dynamics. Field Crops Res., 4: 285-296. Cooke, R.D., Hahn, S.K. and Howland, A.K., 1978. Screening cassava for low cyanide using an enzymatic assay. Exp. Agric., 14: 367-372. Digby, P.G.N., 1979. Modified joint regression analysis for incomplete variety × environment. J. Agric. Sci. Camb., 93: 81-88. Eberhart, S.A. and Russell, W.A., 1966. Stability parameters for comparing varieties. Crop Sci., 6: 36--40. EI-Sharkawy, M., Cock, J.H., Lynam, J.K., Hernandez, A.P. and Cadavid, L.F., 1990. Relasionship between biomass, root yield and single leaf photosynthesis in field grown cassava. Field Crops Res., 25: 183-201.

E1-Sharkawy, M., Hernandez, A.P. and Hershey, C., 1991. Yield stability of cassava during prolonged mid-season water stress. Exp. Agric., 28: 165-174. Finlay, K.W. and Wilkinson, G.N., 1963. The analysis of adaptation in a plant breeding programme. Aust. J. Agric. Res., 14: 742754. Hershey, C.H., 1984. Breeding cassava for adaptation to stress conditions: Development of a methodology. In: CIP (Centro Internacional de la Papa). Proc. 6th. Symp. Int. Soc. Trop. Root Crops. CIP, Lima, Peru, pp. 303-314. Kawano, K., 1990. Harvest index and evolution of major food crop cultivars in the tropics. Euphytica, 46:195-202. Kawano, K., Daza, P., Amaya, A., Rfos, M. and Goncalvez, W.M., 1978. Evaluation of cassava germplasm for productivity. Crop Sci., 18: 377-380. Pellet, D.M., 1992. Physiological aspects of phosphorus response in cassava varieties. PhD thesis, Swiss Federal Institute of Technology, Zurich, Switzerland, 71 pp. Rosielle, A.A. and Hamblin, J., 1981. Theoretical aspects of selection for yield in stress and non-stress environments. Crop Sci., 21: 943-946. Shorter, R., Byth, D.E. and Mungomery, V.E., 1977. Genotype × environment interactions and environmental adaptation. II. Assessment of environmental contributions. Aust. J. Agric. Res., 28: 223-235. Shorter, R., Lawn, R.J. and Hammer, G.L., 1991. Improving genotypic adaptation in crops. A role for breeders, physiologists and modellers. Exp. Agric., 27:155-175. Wholey, D.W. and Cock, J.H., 1974. Onset and rate of root bulking in cassava. Exp. Agric., 10: 193-198. Williams, H.J. and Edwards, T.G., 1980. Estimation of cyanide with alkaline picrate. J. Sci. Food Agric., 31 : 15-22.