I I I I I IIIIIIIIIII'IIIIIIIII’ THESIS IBMRI Illiliillllllllllfilll Illllllll 3 1293 01025 1068 This is to certify that the dissertation entitled Combining Ability among and within early, medium, and late maturity classes of maize (Zea mays L.) inbred lines presented by Mohamed Barre Ahmed has been accepted towards fulfillment of the requirements for Ph.D . Crop and Soil Science degree m Naomi 9024 MS U is an Affirmative Action/Equal Opportunity Institution 0-12771 LIBRARY Michigan State Unlverslty PLACE III RETURN BOXto remove thle checkout from your record. TO AVOID FINES return on or before dete due. DATE DUE DATE DUE DATE DUE a _—_—I .I_——-_|I-—- IF]; MSU le An Nflrmetlve Adlai/Ewe! Opportunlty lmflttflon COMBINING ABILITY AMONG AND WITHIN EARLY, MEDIUM, AND LATE MATURITY CLASSES OF MAIZE (Zea mays L.) INBRED LINES By Mohamed Barre Ahmed A DISSERTATION Submitted to Michigan State University in partial requirements for the degree of DOCTOR OF PHILISOPHY Department of Crop and Soil Sciences 1993 ABSTRACT COMBINING ABILITY AMONG AND WITHH‘I EARLY, DIEDIUM, AND LATE MATURITY CLASSES OF MAIZE (Zea mays L.) INBRED LINES By Mohamed Barre Ahmed Twelve inbred lines consisting of four from each of three maturity classes, were studied. The lines were, early; A641, leOS, W117Ht and M574; medium; W64A, M575, A619 and A632; late; B73Ht, M017Ht, BB4 and Pa872. All possible single- crosses were produced among the twelve lines and evaluated at three locations in Michigan in three years. Data was obtained onlgrain‘yield, stalk lodging, root lodging, moisture content, plant height, ear height and stand count in all experiments. Days to 50% pollen shed and 50% silking were recorded at the Central location in all years. The objectives of this research were to estimate the general and specific combining abilities of lines among and within maturity classes and how these effects were influenced by year and/or location and, to identify the lines with desirable general and/or specific combining ability for earliness and yield. The progeny means were used for combining ability analyses for nine characters, according to a modified version of Griffing's method four, model one. The mean performance of the crosses revealed a wide range of variation among entries for all characters. Maximum range was recorded for stalk lodging (2.6-30.1%). The mean yields of the line crosses showed late lines with the highest values, followed by medium maturity lines, but positive GCA effects were observed for B73Ht, M017Ht, A632, and W64A. A11 early lines gave the lowest cross means, but positive GCA effects were shown, except for M574. The present study pointed out that GCA over all lines, and GCA among and within classes were significant for all characters, whereas SCA over all lines, and SCA among and within classes were significant only for yield, moisture content, plant height, ear height, days to pollen shed and silking date. The additive genetic effects were important for all characters as shown by consistently larger GCA mean squares than those for SCA. The interaction of GCA by year, GCA by location, and GCA by year by location were significant for all characters, except stand counts (LXGCA, LXGCA among and within classes). The interactions of SCA by year were significant for yield, stalk lodging, moisture content and days to pollen shed, while SCA by year by location was significant only for moisture content. The magnitude of GCA by environment was less than those of the main effects, indicating that the additive gene effects were relatively insensitive to the effects of the environments. Three early (A641, Cm105, W117Ht), two medium (W64A, A632) and two late lines (B73Ht, M017Ht) had desirable GCA's for at least four characters, including grain yield. M574 and M575 had desirable GCA's only for maturity-related characters. The line B84 had undesirable GCA effects for all characters. Early lines as a class were low-yielding, and showed lower moisture content, fewer days to pollen shed, earlier silking date, and higher lodging incidence, and produced shorter plants with low ear height, whereas late lines were the complete opposite as shown by their GCA effects. Out of the ten crosses with significant SCA effects, only four had positive values for yield. They were two medium by medium, one medium by late and one late by late. The remaining six crosses had negative values and consisted of two early by medium, one medium by medium, two medium by late and one late by late. To my dear mother, who passed away in 1991, but will always stay with me in spirit. ACKNOWLEDGEIVIENTS I would like to express my sincere gratitude to Dr. Dale Harpstead for his guidance, encouragement and confidence in me during the execution of this project. His fruitful discussions and consultations always seemed to have as a concern my potential to develop and gain research experience which would enable me to become a better scientist. Without his assistance in obtaining financial support from various sources, this dream would have been just a mirage. I also extend my deepest appreciation to Dr. Thomas Isleib for the endless time in assisting me in the computer programs for data processing and in sharing his expertise in the analysis of the diallel cross. A special note should be made of the early contributions of the late Dr. Elmer Rossman who contributed unselfishly during the initial phase of this study. I am thankful to Dr. Wayne Adams, Dr. Daniel Keathley and Dr. Brian Diers who served as committee members and in reviewing this dissertation. A hearty thanks goes to Dr. Carter Harrison for editing my dissertation from its inception to the final form. Likewise, I am deeply indebted to Keith Dysinger and Mervin Chamberlain for their assistence in the field work. I would be remiss if I did not express profound gratitude to Dr. Everett Everson, Dr. Russell Freed, and Dr. Richard Ward VI for their encouragement and assistance in many ways. Finally, I would like to express my deepest and sincere appreciation to my family for enduring my long absence, particularly in this difficult time of the civil war in Somalia and their unfailing encouragement, support and prayer. This degree is as much theirs as mine. VII TABLE OF CONTENTS LIST O!‘ TABLES ......................................... VIII LIUTCMIITGURES ......................................u. XIV INTRODUCTION ........................................... 1 LITERATURE RBVIBI ...................................... 5 Maturity ............................................ s Combining ability ..... 14 Diallel analysis ... 28 MATRIALMANDHMETHODS .................................... 33 Genetic material .................................... 33 Field procedures ... 35 Characters measured 37 statistical analysis . 40 RESULTS ................................................ 58 Grain yield .. 80 Stalk lodging .. 87 Root lodging .. 93 Grain moisture content ... 98 Plant height .... 103 Bar height .. 107 Stand count .. 111 Days to pollen shed and silking date ................ 115 Grifing'snethod .. 123 DISCUSSION ............................................. 124 RESERENCES ............................................. 137 ”Pme......OOOOOOOO.......OOOOQOOCOOOOOOOOO00......O 1“ VIII TABLE 1 10 11 12 LIST OF TABLES Page Pedigree, origin, and release year of the twelve inbred lines used in the diallel ................... 33 Intonation describing the experimental locations . . 38 Fornat of the analysis of variance for single experiment ......................................... 40 Format of the analysis of variance for an individual experimentfordiallelcrosses.....................48 Format of the analysis of variance combined over yearsfordiallelcrosses..........................49 Format of the analysis of variance combined over locationsfordiallelcrosses......................51 Format of the analysis of variance combined over years and locations for diallel crosses ............ 53 Mean of seven agronomic characters of 1989 averaged overthreelocations...............................57 Mean of seven agronomic characters of 1990 averaged overthreelocations...............................57 Mean of seven agronomic characters of 1991 averaged overthreelocations...............................SS Mean of seven agronomic characters of Southern location averaged over three years ................. 58 Mean of seven agronomic characters of Central location averaged over three years . . . . . . . . . . . . . . . . . 59 IX TABLE 13 14 15 16 17 18 19 20 21 Page Mean of seven agronomic characters of Northern location averaged over three years ................. 59 Mean of four agronomic characters averaged over threeyears and three locations .................... 60 Mean of five agronomic characters averaged over threeyears and three locations .................... 62 Mean performance of line crosses for nine agronomic characters averaged over three years and three locations .......................................... 64 Analysis of variance for Grain Yield (t/ha) for 66 diallel crosses evaluated at three locations in 1989, 199031161991...................................... 65 Analysis of variance for Stalk Lodging (t) for 66 diallel crosses evaluated at three locations in 1989, 1990lndl991...................................... 66 Analysis of variance for Root Lodging (%) for 66 diallel crosses evaluated at three locations in 1989, 199011161991...................................... 67 Analysis of variance for Grain Moisture content (%) for 66 diallel crosses evaluated at three locations in 1989, 1990 and 1991 ................................ 68 Analysis of variance for Plant Height (cm) for 66 diallel crosses evaluated at three locations in 1989, 199°.nd19910.0.0..........OOOOOOOOOOOO0.0.00.0... 69 TABLE 23 24 25 26 27 28 29 Page Analysis of variance for Ear Height (cm) for 66 diallel crosses evaluated at three locations in 1989, 1990 and 1991 ................................ 70 Analysis of variance for Stand Count (I of plants/ha) for 66 diallel crosses evaluated at three locations in 1989, 1990 and 1991 ................................ 71 Analysis of variance for Days to pollen Shed (dap) for 66 diallel crosses evaluated at Central location in 1989'199°“d1’91......OOOOOOOOOOOOOOOQII00......72 Analysis of variance for Silking Date (dap) for 66 diallel crosses evaluated at Central locations in 1909, 1990 and 1991 ................................ 73 Estimates of general combining ability effects of Grain Yield (t/ha) over and three years and threelocations....................................79 Estimates of specific combining ability effects of Grain Yield (t/ha) over three years and threelocations....................................80 Estimates of general combining ability effects of Stalk Lodging (t) over three years and threelocations....................................86 Estimates of specific combining ability effects of Stalk Lodging (t) over three years and thr.a1°c‘ti°n8.....OOOOOOOOOOOOOOOOOO0.0.0.0000...87 XI TABLE 30 31 32 33 34 35 36 37 Page Estimates of general combining ability effects of Root Lodging (t) over three locations and three years .............................................. 92 Estimates of specific combining ability effects of Root Lodging (t) over three locations and three years 93 Estimates of general combining ability effects of Moisture Content (t) over three years and threelocations....................................97 Estimates of specific combining ability effects of Moisture Content (%) over three years and threelocations....................................98 Estimates of general combining ability effects of Plant Height (on) over three years and threelocations...................................102 Estimates of specific combining ability effects of Plant Height (on) over three years and threelocations...................................103 Estimates of general combining ability effects of Ear Height (cm) over three years and threelocations...................................106 Estimates of specific combining ability effects of Ear Height (cm) over three years and thr..1°°‘t1°na......OCOOOOOOOIOOI......OOOOOOOOOO 107 XII TABLE 38 39 40 41 42 43 Page Estimates of general combining ability effects of Stand Count (I of plants/ha) over three years and threelocations...................................110 Estimates of specific combining ability effects of Stand Count (# of plants/ha) over three years and threelocations...................................111 Estimates of general combining ability effects of Days to pollen Shed (dap) over three years at Centrallocation..................................115 Estimates of specific combining ability effects of Days to pollen Shed (dap) over three years at Centrallocation..................................116 Estimates of general combining ability effects of Silking Date (dap) over three years at Central location .......................................... 118 Estimates of specific combining ability effects of Silking Date (dap) over three years at Central location ....0.........IOOOOIOOIOIOOOIO0.000.000... 119 XIII LIST OF FIGURES Figure Page 1 The crossing plan of twelve maise inbreds, four from each of three maturity classes (early, medium, .ndl‘t.) I.......OOOOOOOOOCOOOOOOOO......OOOOOOOOOC‘2 XIV INTRODUCTION The expansion of maize (Zea mays L.) production into the short-growing season areas in temperate zones is restricted by low spring temperatures and early fall frosts. The latter may terminate growth prior to physiological maturity and consequently decrease the effective growing season. In the tropics, particularly in semi-arid regions, successful maize production is constantly threatened by limited rainfall and lack of supplementary irrigation. Because of these effects, the most desirable maize genotypes may be those which can partially evade drought by having a shorter vegetative cycle. Maturity is therefore an essential attribute to consider when choosing maize hybrids or populations for use in regions with short growing seasons. Recently intensified efforts have been exerted to study characteristics which constitute the components of earliness. These are number of effective growing days from planting to flowering, the length of the effective grain filling period and moisture content at harvest. Early maturing plant genotypes, however, have frequently been associated with reduced yield potential even when growing conditions are favorable. The reason for lower grain yields has been attributed to the inability of a genotype to take full advantage of the available growing season (Hallauer et al., zones has often resulted in an increase in the maturity range 1967). Selection for yield improvement pg; 5; in temperate of the material (Hallauer and Miranda, 1981). Attempts have been made to reduce vegetative period and increase yield by selecting for the two characters simultaneously. Unfortunately, the breeder faces considerable difficulties in assigning subjective economic values or proper weights to the relative importance of each trait. A second problem exists in judging with confidence the best parents or crosses in the breeding program. The presence of significant heterotic responses in crosses among unrelated populations of maize has been known for a long time and has stimulated the exploitation of genetic diversity. A search for efficient utilization of diversity in maize germplasm has created renewed interest in searching for heterotic responses based on geographical separation, climate adaptation, endosperm color, kernel texture, divergent selection, and, finally, maturity. Diallel analysis has been used extensively to evaluate parents for their combining ability. Information on the relative size of general and specific combining ability effects is used to facilitate the selection of parents and to develop crossing plans that offer the best promise of obtaining elitehybrids or superior composites. A widely accepted concept among maize breeders is that crosses between different maturity classes of maize often result in higher general combining ability (GCA) effects than crosses within maturity classes [Cross, 1991; Moll, 1991; Goodman, 1991; Hallauer (personal communication, 1991)]. However, a search of the literature failed to uncover comprehensive studies dealing directly with combining ability among and within different maturity classes. In general, maize breeders use diallel analyses to evaluate the combining ability of their advanced.breeding lines without reporting the relative maturity of the lines. In addition, in most of the cases where maturity has been reported, the lines were either in a single maturity class or were of different maturity classes, but treated together as one group. In the latter case, total variances for general and specific combining ability were not partitioned among and within maturity classes (Bhala et al., 1977). Occasionally, when the lines were from various maturity classes they were divided into different diallel sets according to maturity (Cross 1977, Han 1991). Information on the combining ability among and within classes is of prime interest in progeny evaluation. Combining ability measurements among and within maturity classes were identified as an essential objective of this study. A twelve inbred line diallel set was created from four early maturity lines, four medium lines, and four late lines. The objectives of this research were to: 1) obtain estimates of general and specific combining ability among and within maturity classes, 2) determine the effect of location and year on GCA and SCA effects, and 3) identify the best combination for earliness and yield in terms of combining ability. LITERATURE REVIEW MATURITY Early maturity is very important in the short growing season areas of North America where frost damage is likely to occur before the crop reaches physiological maturity. Frost damage results in significant absolute loss of yield and greatly increases cost of production in terms of drying and storage problems. Moreover, harvesting may become difficult during inclement fall weather. Attention has been directed to the determination of what makes one variety or hybrid mature and ready for harvest sooner than another. Time to maturity is a complex trait that is difficult to assess. Various measures . have been proposed to characterize relative maturity. Many of these estimates may be based on external appearances of the plant while others may rely on internal measurements made on either the tassel, the ear or the whole plant. Aldrich (1943) defined maturity (or readiness for harvest) as the point at which the maximum dry weight of the grain is first attained. This has also been referred to as "physiological" or "morphological maturity" (Shaw and Loomis, 1950; Anderson, 1955). Grain moisture content at harvest has become an accepted criterion of maturity, but its total acceptance has been eroded because of the wide ranges that have been associated with maximum dry matter accumulation and the failure to identify the rate at which dry down was taking place. There has been a growing dissatisfaction with the lack of a precise definition and consistent use of a given maturity index in order to obtain an accurate evaluation of breeding material for a proper maturity rating. Early work in Iowa indicated that the interval from silking to maturity appeared to be influenced little by environmental conditions with no significant differences between early and late varieties. Therefore, the time of maturity has frequently been predicted on the basis of the date of mid-silking to the date where physiological maturity occurred (Shaw and Thom, 1951). Studies of the inheritance of maturity in maize have focused on the number of days from planting to mid-flowering. Jones (1955), using six early by late crosses of maize inbred lines, postulated that the minimum number of genes affecting date of silking ranged from five to nineteen. This work suggested that earliness was mainly due to dominant genes, with complete dominance for early silking. Giesbrecht (1960) calculated a low heritability estimate for flowering date, and indicated that maturity was controlled by more than two genes with some evidence of partial to complete dominance and epistasis effects. In another study involving genetic analysis of maize inbreds and the progeny of their crosses, Giesbrecht (1960) identified the existence of partial phenotypic dominance for earliness and interallelic interactions of maturity factors. In a study of two inbred lines and their hybrid combinations, Hallauer and Russell (1962) defined maturity as the average time at which the maximum dry weight of the grain was attained, with maturity measured as the number of days from silking to that point. Although the lines were deliberately selected for their differences in date of silking and grain moisture at harvest, they reached maturity at approximately the same time (60 days) from silking to a grain moisture of 36.4%. These authors suggested that grain moisture alone could not be considered to be a true indicator of maturity for specific material. In the cases studied, the interval from silking to maturity was found to be relatively constant within years. Estimates of various components of genetic variance for grain moisture and kernel weight were obtained from six generation means. The results indicated that the largest portion of genetic variance was attributed to dominance effects, but the estimates may be confounded with epistatic effects. Hallauer (1965) examined the inheritance of time to flowering of the inbred lines B14 and 0h45, and their single crosses. He found that the estimate of additive (D) genetic variance was much greater than the dominance (H) variance. However, he pointed out that linkage might have biased the estimate of additive variance. Based on the idea that various maturity groups of maize genotypes require different amounts of accumulated thermal energy to reach maturity, interest has been shown in expressing maturity in terms of accumulated thermal units (degree days) as a measure of the growing season. Gilmore and Rogers (1958) compared fifteen different methods of calculating heat units required for silking. They suggested that the common methods used for computing degree days (i. e., x -10°C) could be improved by correcting for the effect of low or high temperatures on growth. They concluded that the maturity could be more accurately determined by calculating the degree days or "effective degrees" required to bring genotypes to silking than by calendar days. A significant relationship between leaf number and maturity has been reported in maize (Chase and Nanda, 1967; Arnold, 1969; Allen et al., 1973; Ahmed, 1987). Evidence presented by Chase and Nanda revealed the existence of a highly significant correlation between leaf number and maturity, with late-maturity hybrids having more leaves than early-maturity hybrids. They suggested that the number of leaves characterizing maize hybrids provides a basis for maturity classification. However, leaf number is influenced by environmental variables and growing conditions (Allen et al., 1973). It has been observed that increasing temperature (over certain ranges) and high fertility tend to result in a slight increase in leaf number, whereas, increasing plant population may result in a decrease of leaf number. Increasing photoperiod may have a major effect on increasing leaf number. Bonaparte (1977) investigated the inheritance of leaf number and.maturity as determined by the number of days from planting to mid-silking in a diallel cross. He used six inbred lines and their single cross hybrids which had differences in maturity and leaf number. He found that days to mid-silk and leaf number were controlled by at least one and four factors, respectively. Heritability estimates in both a broad sense and narrow sense for these characters were high. Grain yield is the result of dry matter accumulation in the grain during the period from silk emergence to maximum kernel dry weight (grain filling period). Discrepancies exist in the literature regarding the length of this period. Shaw and Thom (1951) obtained a relatively constant time interval of about 51 days for three hybrids. In contrast, Dessureaux et al., (1948), found differences of at least seven days among four inbreds in the duration of the grain filling period. Recently, several workers reported sizable differences (eight days or more) among maize genotypes in the duration of this period (Carter and Poneleit 1973; Daynard et al., 1971; Daynard and Kanennberg 1976; Gunn and Christensen 1965). However, it is important to know when the kernels have ceased starch accumulation in order to determine the duration of 10 filling period. Daynard and Duncan (1969) suggested that black layer formation at the kernel base was a good indicator of when total dry'matter accumulation had.been.reached, and it is now accepted that the appearance of this black abscision layer signals the physiological maturity of the kernels. In this method, the termination of filling period is monitored by sampling kernels from the middle of the ear, because the kernels in the tip region tend to reach the black layer rather earlier than those from the rest of the ear. In attempts to understand the differences for grain yield among various maturity classes of maize hybrids, considerable attention has been given to the rate of grain filling, as well as the duration of grain filling. Previous research reports indicated that a positive relationship exists between yield and the duration of grain filling. Gunn and Christensen (1965) found that later maturing hybrids had a longer filling period and larger kernels than earlier maturing hybrids. Daynard et al., (1971) observed that yield differences among hybrids could be more accurately described in terms of an effective filling period duration. Several hypotheses have been presented pertaining to mechanisms by which maize genotypes may regulate the duration of the grain filling period. Duncan (1975) extrapolated that one possible determinant of the duration.of the filling period could be the relationship between photosynthetic rate and sink capacity (i. 11 e., where sink capacity is defined as the number of kernels per plant times the capacity of these kernels to accommodate assimilates) . Others have speculated that kernel volume may be primarily responsible for the establishment of the duration of the filling period (Daynard and Kanennberg, 1976). Tollenaar and Daynard (1978) studied kernel growth and development and suggested that kernel strength ( i.e., where sink capacity is referred as the capacity to attract assimilates) may affect the rate of dry matter accumulation. Enhancement of photosynthetic supply during flowering and grain filling period may increase the rate of dry matter accumulation. This could result in the formation of larger kernels which contribute to improved grain yield in the short growing season areas of North America. Furthermore, Mock and Pearce (1975) included a longer grain filling period as a major component for the development of a maize ideotype. The authors suggested that this period, however, should not be so long that leaf senescence occurs before maximum grain dry matter is attained. In the northern Corn Belt, where the growing season is short, leaf senescence often does not occur prior to the accumulation of maximum dry matter in the grain. A higher rate of kernel development could lead to an advanced harvest date which could be of agronomic significance in that region. Carter and Poneleit (1973) observed that some inbred lines with similar duration of filling period, had differences 12 in the duration of the vegetative phase as well as the period from planting to physiological maturity. Hillson and Penny (1965) studied the date of silking and the number of days from silking to physiological maturity as a measure of maturity. They found that some hybrids which silked earlier required more time for the filling period than some later silking hybrids. Other researchers have also found that some exceptional hybrids with a short filling period had attained a considerable kernel weight in a relatively short time as a result of a very high rate of grain dry matter accumulation (Daynard and Kannenberg 1976; Perenzin et al., 1980). Recent evidence suggests the presence of substantial genetic potential for grain yield improvement through either extension of the filling period (Daynard et al., 1971) and/or enhancement of kernel growth rate (Tollenaar 1977). These traits appear to have a high heritability (Perenzin et al., 1980) and are controlled primarily by genes with additive effects (Cross, 1975). However, the potential yield increase through extension of the filling period in the Northern latitudes is limited by the length of the frost free interval. Cross (1975) used a diallel cross of seven inbreds to examine the duration and rate of filling and to determine the possibility of selection for these traits in breeding programs. He reported that general combining ability effects for these traits were larger than specific combining ability 13 effects, indicating that simple selection procedures which take advantage of additive genetic variances should be effective in changing these traits. Unfortunately, the practical application of selection for these traits in a breeding program is often hindered by a lack of a simple and inexpensive screening method(s), especially for evaluation of a larger number of genotypes. COMBINING ABILITY The ability of a parent to transmit desirable performance to its hybrid progeny has been referred to as combining ability. Sprague and Tatum (1942) first introduced the partition of combining ability into general and specific effects. They defined general combining ability (GCA) as the average performance of a parent in hybrid combination with all other parents in a population, and interpreted its variance as a measure of the additive genetic portion. Specific combining ability (SCA) was designated as the deviation from the expected performance of a cross based on the average performance of its parents. The variance of SCA is interpreted as a measure of the non-additive genetic portion, and includes dominance and epistatic effects. The method of evaluation for combining ability and the choice of material included in the test has been changed from topcross to diallel cross (Hallauer and Miranda, 1981). Diallel cross helps you to determine not only the average performance of a variety or a line in comparison with other varieties, but also a nicking of a particular pair of varieties. When specific crosses are important to the breeder, the diallel design provides more directly useful information related to that end than using random testers as in topcross procedure (Sprague, 1984) . The relative importance of GCA and SCA in maize has been studied by numerous 14 15 researchers. Sprague and Tatum (1942) utilized the diallel cross to evaluate the GCA and SCA of unselected lines. They found that GCA was more important than SCA in controlling grain yield. When they used single crosses that had been previously selected for general combining ability however, they found that the variance of SCA was greater than that of GCA. Rojas and Sprague (1952) obtained similar results from experiments involving two groups of previously selected material. The variance of non-additive genetic effects was consistently greater than the variance of additive genetic effects. The authors also extended the analysis for general and specific combining ability to include interactions among locations and years. They reported that the interaction components involving SCA was also greater than that for GCA, indicating that genotype by environment interaction may be an important contributing factor to the variance of SCA. Matzinger (1952) later took the diallel cross analysis procedure one step further, by outlining the models for estimation of genetic variances for experiments repeated over years and locations. Matzinger et al., (1959) reported yields of 10 unselected lines from a maize synthetic variety evaluated at three locations for three years. A.dialle1 set of crosses was used to compare the variances of GCA and SCA and their interactions with years and locations. When the data were combined over years and locations, the GCA variance 16 components were decreased, suggesting a strong influence of SCA effects in developing stable yield over environments. They discussed estimates of general and specific combining ability more completely with respect to their expectations in terms of additive and dominance genetic effects. Hallauer (1971) reached conclusions similar to those of Rojas and Sprague (1952), and Matzinger et al., (1959). He pointed out that substantial genetic variance was lost to the non-additive genetic component when grain yield from two Iowa synthetic cultivars were combined over years and locations. Estimates of GCA and SCA have been used by plant breeders to make decisions pertaining to the most suitable breeding method(s) and in selecting parents for breeding programs. The ratio of mean squares of GCA and SCA is used to determine which one of the two is more important. The prevalence of GCA facilitates the selection of parents to use for intrapopulation improvement. Furthermore, the presence of larger GCA also increases the chances of obtaining a superior synthetic variety because non-additive genetic variance diminishes in advanced generations of random mating. In, contrast, a SCA of greater magnitude would help establish heterotic patterns among inbreds for potential hybrids or among parents selected for interpopulation improvement where divergent populations may eventually be used as a source of inbred lines. 17 The presence of a substantial quantity of genetic variability for most agronomic characters is a key factor in choosing source germplasm in any breeding program. The level of total genetic variability and the nature of gene action in the source germplasm will ultimately determine the optimal breeding methods, efficiency of selection, and final success. The level of genetic variability can be improved by increasing genetic diversity (Lonnquist, 1953; Lonnquist et al., 1961). It has been shown that materials of diverse genetic origin produce better hybrids than those more closely related (Moll et al., 1962; Moll et al., 1965; Paterniani et al., 1963). Genetic diversity has been associated with increased heterosis, which is an important factor in utilizing germplasm to maximize the cross-performance of source populations. Maize breeders have, therefore, devoted considerable effort and time to the development not only of genetically variable source populations, but also populations with unique heterotic patterns. Such source populations may be obtained in one of several possible ways; for example, by using populations of different geographic origin, of contrasting endosperm types, or of widely varying maturity. The early literature describing maize improvement attempts indicate that crosses between varieties with contrasting endosperm type (dent x floury and dent x flint) were more likely to outyield the parental varieties than 18 crosses between varieties from the same kernel texture group (Jenkins, 1978). A striking example was provided by the crosses of early Northern Flints with late Southern Dents which generated many excellent open-pollinated varieties such as Reid Yellow Dent. Another example is the development of Lancaster Surecrop in Pennsylvania which probably came from crosses between early flint and a late local variety. Early studies recognized that the crosses between the lines from Reid Yellow Dent and Lancaster Surecrop provided excellent hybrid combinations. Consequently, these two varieties have formed the basis of the most widely-used heterotic pattern in maize breeding, and have resulted in the most extensively used lines in temperate hybrids, especially in the U.S. cornbelt (Jenkins, 1978). Wellhausen (1978) pointed out that crosses between Tuxpeno and its related Caribbean and US dents with Cuban flints and Coastal Tropical flints often exhibit a high level of population-cross performance. He suggested the development of two broad-based, high yielding populations consisting, respectively, of a dent and a flint composite. The dent composite could be from Tuxpeno and related dents, while the flint composite could be from the Cuban, Coastal Tropical and Cateto material. This recommendation stimulated the study of population crosses between dent and flint types with excellent heterotic patterns in different parts of the world. In the 19 tropics several heterotic groups have been identified and used extensively to develop hybrids. The most important heterotic combinations are Tuxpeno by Cuban flint, Tuxpeno by Coastal Tropical Flints, and Tuxpeno by ETO (flint). In Kenya the heterotic pattern involving a Tuxpeno-derived Kitale selection (Kitale II) and a high altitude flint material from Ecuador (EC573) has been exploited for the East African highlands. Crosses between early European flints and US Corn Belt dents have been successfully utilized in Europe (Eberhart et al., 1991). Moreno-Gonzalez (1988) produced a diallel cross of seven flint (F) and seven dent (D) inbred lines of maize. He divided the crosses into three subdiallel sets, namely FXF, DXD, FXD and partitioned the variance of each of the first two groups into that due to GCA and SCA, following the Griffing method four (1956). In the FXD subdiallel, he used the Comstock et al., (1952) design to partition the total variance into that for GCA of each F and D group and that for SCA of FXD. He found that the GCA effects for yield were higher in dent than in flint lines. The crosses of FXD predominantly had positive SCA effects but negative SCA effects were observed in FXF and DXD crosses. Similarly, Dhillon et al., (1979) presented data of crosses between five flint and five dent inbred lines of maize. An incomplete factorial design was employed to adjust for missing crosses. They followed the procedures presented by 20 Milliken et al., (1970) for thetanalysis of combining ability. These procedures partition the total variance into the variance of GCA for flint and dent lines. They found a significant mean square for GCA of both flint and dent lines for grain yield. However, they also reported that the interaction between GCA and environments for the dent lines was not significant, indicating that the dent lines were adapted to the region. The superiority of crosses between flint and dent varieties have stimulated a search for a possible maximum benefit as well as an efficient utilization of diversity as reflected in differences of geographical separation and/or adaptation. A diallel study involving three Corn Belt synthetics and three exotic synthetics has shown positive GCA effects for the first group and negative effects for the second (Gerrish, 1983). This could be explained only in part to a lower frequency of favorable alleles for yield in the exotics. In fact, crosses within each group in the study showed that Corn Belt dents yielded substantially higher than their counterparts, and crosses among exotic groups were intermediate by comparison. Nevertheless, significant GCA has been observed which might indicate the consistent contribution of the synthetics to their crosses. This finding was in good agreement with what would have been expected if additive gene action was of prime importance. It is interesting to note that 21 the best combiners in this study showed similar performance for grain yield in all environments (Gerrish, 1983). In India, ten varieties were selected for differences of origin as well as for chromosome knob number and crossed in all possible single cross combinations. Estimates of GCA for grain yield were studied and accordingly the parental varieties were placed into three classes, high, average, and low combiners. Evaluation of the relationship between GCA and yield and knob number for the parents indicated that high and average combiners had high knob numbers, while low combining varieties always possessed low chromosome knob numbers (Kalsy et al., 1970) Utilization of this close association of high knob number with high GCA of the parents and the possibility of using knob number for determining the heterotic effects between crosses of the parents was discussed. In a study based on an unrelated genetic relationship, Rinke and Hayes (1964) selected fifteen inbred lines and classified them into inferior (I), medium (M), and superior (S) GCA classes. The yield for each of the three I, M, and 8, classes ranged from 93-103, 104-106, 110-113 % of the check, respectively. They concluded that according to the frequency distribution of the FI performances, the highest percentage of outstanding hybrids was most often derived from crosses of S by S GCA lines. They did, however, acknowledge the presence of one exceptional cross of M by S which gave a similar 22 performance to those of S by 8 lines. Beck et a1. , (1991) evaluated the performance of diallel crosses among nine of CIMMYT's subtropical and temperate intermediate-maturity maize gene pools and populations. The material was tested at five locations in Mexico and eleven in the USA. They reported highly significant and positive GCA effects for yield with Populations 42, 47, and 34 in Mexican environments. Only Pool 41, which had more temperate germplasm, exhibited positive GCA effects for yield in the US environments. The combining ability of eight parents as measured in their crosses was reported by Everett et al., (1987). These materials‘were divided into high- (3), mid- (1), and low-altitude (4) genotypes based on their known adaptation. Evaluation was carried out at four locations ranging in altitude from 1000 to 1600 meters in Cameroon. They found that the lowland material combined better with highland materials for yield, plant type, and resistance to lodging. The lowland populations were short in plant stature and probably early in maturity, whereas the highland populations were tall and late in maturity. The authors, however, did not indicate whether maturity had any influence on combining ability of the material. The results also show that the mid- elevation population (MSR) combined better with highland than with lowland populations in terms of yield and ear quality. Diallel analysis has also been used to evaluate the 23 improvement that has been made through several cycles of recurrent selection. Stangland et al., (1983) evaluated crosses of four populations and four selected 82 lines from each population for yield and other agronomic characters. They partitioned the total variation for GCA and SCA into variation due to population and variation attributed to the lines within a population. The GCA accounted for most of the genetic variation of the populations and the lines, but the corresponding SCA.was also highly significant for all traits. A considerable portion of the variance for both GCA and SCA for yield, however, was accounted for in the populations. The majority of the lines within each population exhibited GCA effects similar in magnitude and direction to their population sources. Only the SCA for the lines and GCA for both populations and lines showed highly significant interactions with environments. In contrast, the SCA for the parental sources were not only non-significant but also very small relative to the SCA for the lines. The interaction variances involving GCA were much larger in magnitude than those involving SCA. There has been a growing interest with regard to the combining ability of material developed through bidirectional selection. Crosses of the lines developed from these divergent populations may provide information of the underlying genetic mechanism and produce a more desirable result. The combining 24 ability expressed in High X High (HH), High X Low (HL), and Low x Low (LL) single crosses of maize inbred lines from the Krug variety, obtained through divergent selection for yield in several generations, was studied by Lonnquist (1953). He indicated that the greatest possibility of obtaining a considerable number of high-yielding crosses was from crossing HH combining lines rather than either HL, or LL groups. The results of his study seemed to support the importance of dominant or partially dominant favorable gene effects in determining hybrid vigor. An analogous experiment was reported by Lamkey and Hallauer (1986). They produced three groups of crosses from 24 high (H) and 24 low (L) lines selected for yield.pg;H§g from 247 random inbred lines of 'Iowa Stiff Stalk Synthetic' (BSSS). Single crosses were produced among and within groups. Significant differences among the three hybrid group means (HH, HL, LL) for grain yield were reported. The ranking of the hybrids (HH>HL>LL) was as expected under a model with partial-to-complete dominance. They concluded that GCA was more important for the variation observed than SCA, which indicated that additive type gene action was predominant. Although maize breeders generally accept the premise that crosses among different maturity classes results in higher GCA effects than crosses within classes, empirical results in this area are lacking in the literature [Cross, 25 1991; Moll, 1991; Goodman, 1991; Hallauer, 1991,(personal communication) ] . Diallel analysis has frequently been used to evaluate the combining ability of advanced breeding lines without reporting the relative maturity of the lines tested. In addition, in most of the cases where the maturity was reported, the lines were either in a single maturity class or were in different maturity classes but.treated together as one group. In the latter case, the total variances for general and specific combining ability were not partitioned among and within maturity classes. Crosses of five early and five late maturing open-pollinated varieties has been reported by Bhalla et al., (1977). The material was selected on the basis of plant height (i.e. early were short and late were tall). The GCA effects pooled over environments indicated that four of the five late varieties were the best combiners for yield. However, more of the SCA was accounted for in crosses involving varieties from early by late groups, suggesting that these effects appeared to be greatly influenced by genetic diversity. Combining abilities for all characters revealed major interactions with locations, although the GCA was much larger than SCA. Nevado et al., (1989) reported a study of three diallel sets among eight synthetics by crossing early by late, medium by medium, and late by early flowering plants. They concluded that all groups had similar estimates of GCA and SCA for eight of the nine traits tested. Furthermore, the 26 GCA mean squares were of a larger magnitude than SCA mean squares for all traits, suggesting a predominance of additive genetic variance. Results from a diallel cross among ten tropical early and intermediate gene pools and populations were reported by Beck et al., (1990). The materials were evaluated at nine locations in Mexico and other countries. Estimates of combining ability effects were obtained using Gardner and Eberhart's method III (1966) . They found that the mean squares for GCA_wereihighly significant for grain yield, days to silk, plant and ear height, while only ear height had a significant SCA. The interaction component of combining abilities and location indicated that the SCA's were relatively more stable than GCA's. Furthermore, a comparison of variation due to GCA and SCA showed that the additive genetic effects were predominant for all traits. It is important to note that all early maturing parents exhibited highly significant negative GCA effects for all traits. This indicated that these parents contributed not only earliness, but also a decrease of plant and ear height to their crosses. These effects were also accompanied by a reduction in grain yield. Occasionally, when the lines have been drawn from various maturity classes they have been divided into different diallel sets by maturity. Cross (1977) tested two separate sets of diallel crosses involving a set of early and a set of 27 late lines. He found that GCA in both sets was larger than SCA, and that the interaction of GCA with environments was significant. Han et al., (1991) used 58 lines selected on the basis of their overall performance and divided them into six diallel sets. The lines were derived from 11 CIMMYT intermediate and late maturity gene pools and populations. Each diallel set contained only the lines from one maturity class, except diallel number 5 in which the lines were in both maturity classes. The SCA effects of each diallel were partitioned into effects due to intra- and interpopulation crosses to compare the cross performance of the lines from different populations. In each diallel the authors reported both significant negative and positive GCA effects for the lines derived from each of the parental pools and populations. The result suggested that the GCA effect of a line was not closely associated with the population involved. Furthermore, the SCA effects of the crosses produced from interpopulation lines had a positive sign whereas the intrapopulation line crosses had a negative sign. This would suggest that interpopulation crosses on the average had greater heterosis than intrapopulation line crosses. Vasal et al., (1986) reported the combining ability of two separate diallel sets of CIMMYT's subtropical maize germplasm. The sets consisted of seven early and nine 28 intermediate materials, and were evaluated at several locations in cooperating countries. They found that the GCA effects of early material were significant, but the SCA effects were not significant. However, they indicated that most of the best hybrid combinations were from crosses between materials of contrasting endosperm type or color (i. e. , flint by dent and/or yellow by white). In the second diallel they reported that Population 42 had highly significant GCA effects and also the best performance in hybrid combination with Population 47. DIALLEL ANALYSIS Diallel crosses have been used extensively to obtain information concerning the inheritance of quantitative traits. The procedure involves crossing a set of N genotypes in all possible N2 combinations. Genotypes may be individuals, clones, inbred lines, or open-pollinated varieties. Certain assumptions must be satisfied if this technique is to be considered appropriate for the genetic interpretation for a given data set. Some of the underlying assumptions are: 1. Diploid segregation in the parents, 2. Homozygous parents, 3. No difference between reciprocal crosses, 4. Genes are independently distributed between parents, and 5. No non-allelic interaction (Hayman 1954). 29 Numerous authors have examined the importance of these assumptions for a valid interpretation of diallel experiments and their utilization in a practical breeding program (Kepmthorne, 1956; Gilbert, 1958; Matzinger, and Cockerham, 1963; Sprague, 1966). They have agreed that some assumptions (4 and 5) are more critical than others and that it seems to be unrealistic to impose all of the assumptions in a practical situation. It is safe to assume, however, in the present study, the use of maize inbred lines fulfills the first three assumptions and that the remaining two, based on published experiments of similar materials, have been satisfied. Several statistical techniques are available to the breeder for analyzing the data of a diallel experiment. The use of a particular technique depends on three basic factors: 1. the nature of the material under investigation, 2. the genetic hypothesis, and 3. the method of estimation. Generally, four lines of approach may be followed for diallel analyses (Baker, 1984). 1. Jinks and hayman's method (1953), 2. Kempthorne's method (1956), 3. Griffing's methods (1956), and 4. Gardner and Eberhart's methods (1966), Hayman (1954a,b) has developed an analysis which is based on the biometrical analysis of quantitative variations presented by Mather, 1949. The method is useful for the study 30 of genetic differences of fixed samples of inbred lines. The parameters in the analysis are the means and variances in generations derived from crosses of homozygous parents to get estimates of additive, dominance and epistasis gene effects. The method permits an interpretation of the regression of the array covariance {the covariance between all the offspring of the r... parent and their non-recurrent parents (W,)} on the array variance {the variance of these offspring (“9} which provide a graphical representation of the degree of dominance. It also offers the possibility of separating true dominance from spurious dominance caused by many types of non-allelic interaction. With dominance, the regression line has a slope of one. If the line has a slope of one and passes through the origin, the dominance is complete (H,=D) . Movement of the line with b=1 above or below the origin would reveal a decrease (H,D) of dominance, respectively. In the absence of dominance (Hfidn the points would cluster about the position where there is no significant regression line passing through the origin. In 1956, Griffing presented the analysis of four different diallel crossing systems and explained in detail their relationship with the concept of combining ability. These methods are: 1. All N2 combinations, 2. 1/2 N(N+1)combinations, 31 3. N(N-l) combinations, and 4. 1/2 N(N-l) combinations. The fundamental difference among these analyses is based on the genetic material included in the experiment. If the material consists of parental lines, F3, and reciprocal F, progeny, the diallel is analyzed as Method one. When the F, reciprocals are excluded in the analysis Method two is utilized. Method three contains the F, and F, reciprocals, while Method four consists of only the F, progeny. Griffing (1956) emphasized the significance of sampling procedure used to draw the parental lines. He pointed out that a breeder may choose a set of parental lines for the diallel and may be interested in the individual desirability of the material. Thus, any inferences from the results are limited to this particular set (fixed effects) of parents. On the other hand, a researcher may be concerned with the population in which the parents are considered a random sample and the inferences are applied to a larger population (random effects). From these sampling procedures, two different genetic models, Model one, and Model two were derived resulting in eight different analysis (4 methods and 2 models in each). Most of the time the parents are excluded from the analysis, because of the low yields of inbred lines compared to the hybrids produced from them. Thus, method 4, model 1 has been used extensively for the analysis of combining ability of 32 maize. However, from Griffing's methods, it is possible to estimate the general (g,) and specific (5,) combining ability effects for traits and identify the parents whose hybrid combinations are likely to give the maximum selection potential. In 1966, Gardner and Eberhart developed three methods of analysis for diallel crosses. The authors pointed out that their methods are appropriate only when the parental lines comprise the entire population. They also suggested that most breeding materials in which breeders are interested have been highly selected in favor of economically important traits. Thus, estimation of variance components does not provide useful information if such materials are regarded as a random sample from a large base population. Gardner and Eberhart's methods are applicable not only for inbred lines, but also for open-pollinated varieties. Their analysis II is identical to those of Hayman and of Griffing's method 2, model 1), except that Gardner and Eberhart subdivided the specific combining ability into three levels of heterosis [average(h9, variety(h,), and specific heterosis(h,,)]. Analysis III of Gardner and Eberhart considers only F, progeny and is similar to Griffing's method 4, model 1, and that used by Sprague and Tatum in 1942. INLAITHRLAI.AUNL>LIEHHHCHDS CHHNEHICXNLATTHUUXLS The genetic material investigated in this study consisted of twelve inbred lines, four selected from each of three maturity classes (early, medium, and late season). The lines are listed according to their maturity class, the first four are early, followed by medium and then late, as follows: A641, Cm105, W117Ht, M574, W64A, M575, A619, A632, B73Ht, M017Ht, B84, Pa872. The pedigree, origin, and year of release are listed in Table 1 (Henderson, 1980). Diallel crosses among the twelve lines were obtained in 1988. Crosses were produced by making paired row crosses. The materials were sown at time intervals to facilitate crossing between the different maturity classes. The lines were increased by sib-pollination within two rows. All pollinated ears in a row were harvested by hand, and bulked for each cross. The ears were dried to 15.5% grain moisture, shelled in bulk, and thoroughly mixed before a sample of seed was taken for evaluation trials. In 1989, the F2 generation was obtained by sib- pollinating within two rows of F, hybrids. At the same time, the previously described crossing procedure was repeated to jproduce additional seed of parental lines and.F} hybrids. 33 34 Table 1. Pedigree, origin, and release year of the twelve inbred lines used in the diallel. Inbred Pedigree Origin Year line A641 (NDZO3XBI4) Minnesota 1966 Cm105 (V3XB14) Canada, Manitoba 1970 W117Ht (643XMinn#13) Wisconsin 1964 M574 (Michigan Early Syn.) Michigan 1979 W64A (WF9X187-2) Wisconsin 1956 Ms75 [M5153X8D10)Msl42R] [(SDlOXB37)8670] Michigan 1985 A619 (A171X0h43)0h43 Minnesota 1961 A632 (Mt42XB14)B14 Minnesota 1964 B73Ht (Iowa Stiff Stalk Syn.) Iowa 1958 M017Ht (187-2XC103) Missouri 1964 334 (355513(52))co Iowa 1978 Pa872 (75F-5XPa881P) Pennsylvania 1979 pa EX la Ce Ce: FIELD PROCEDURES Yield trials were conducted at three locations in Michigan in 1989, 1990, and 1991 for a total of nine experiments. The three locations were the Crop and Soil Sciences Research Farm near East Lansing, the Michigan State University Potato Research Farm near Stanton, the Dave and Mel Cripe Farm near Cassopolis, and the Robert and August Oshe Farm near Custer. Cassopolis is located in the Southern (Cass County) part of the State, East Lansing is located in the Central (Ingham County), and Custer (Mason County) and Stanton (Montcalm County) are located in the Northern area. Hereafter, these locations will be referred to as the Southern, Central, and Northern locations respectively. The soil types and.other information related to these experimental locations are presented in Table 2. Entries in each experiment of 1989 consisted of 12 parental lines, the 66 F, hybrids, and 3 commercial check hybrids which represented the three maturity classes. The experimental design at each location was a 9 by 9 simple lattice with three replications. A single row plot 9.12 m (30 feet) long was used in Central, and 11.25 m (37 feet) long in Southern and Northern locations. The rows were spaced 91.4 cm (36 inches) in Central, and 76.2 cm (30 inches) in other locations. The plots 35 36 were overplanted and.hand-thinned.in the 8'to 10 leaf stage to a maximum of 50 plants per plot in all experiments. This gave a desired population density of close to 59,957 plants/ha in Central and 58,336 plants/ha in the other locations. During the 1990 season, in addition to the previous material, 66 F2 generation and 22 commercial checks were included for the evaluation trials. In that year, the Stanton location was substituted for Custer. The experiments were arranged as a 13 by 13 simple lattice design replicated three times in each location. The experimental unit in all locations was a two-row plot of 6.08 m (20 feet) in length, with a distance of 76.2 cm (30 inches) between rows. Each plot was overplanted and later thinned to a desired population density. Because of the differences in vigor associated with the inbreds and.ryigenerations, it was not possible to obtain a yield measurement from the inbreds in the first two test seasons (1989 and 1990). Therefore, it was necessary to use a blocking procedure to eliminate severe competition effects. In 1991 a separate randomized complete block.design with three replications was utilized for the inbreds in all locations. A.12 by 12 simple lattice design with three replications was employed for the evaluation of 66 entries in each of F, and F2 generations, and 12 commercial hybrids. A two-row plot of 6.68 m (22 feet) long and between-row width of 0.76 m (30 37 inches) was used. The plots were overplanted and thinned to a maximum of 60 plants/plot in the Central, and 72 plants/plot in other locations to give the desired number of 58,867 and 70,640 plants/ha respectively. All nine experiments were machine-planted and harvested without gleaning the lodged plants. Recommended agronomic practices, including high fertility and.weed control, were followed at all locations to promote high productivity. Supplemental irrigation was applied to all experiments at the Southern location consistent with the commercial farm practice where the test was located. CHARACTERS MEASURED Data were collected for the following nine agronomic characters. Stand count (SC): Number of plants per plot at harvest were recorded and expressed as the number of plants per hectare. Stalk lodging (SL): The percentage of plants in a given plot broken below the top ear were determined. Root lodging (RL): The percentage of plants in a plot leaning 30 degrees or more from the vertical were determined. Plant height (PH): measured to the nearest centimeter after anthesis as the distance from the ground to the flag leaf collar. Ear height (EH): measured in centimeter as the distance from 38 the ground to the uppermost ear-bearing node. Five competitive plants were used for both plant and ear height determinations. The mean value of the five plants was used for the analysis of variance. Days to pollen shed (DP): The number of days from planting to the day approximately 50% of the plants in a given plot were shedding pollen. Silking date (SD): as the number of days from planting until 50% of the plants in a plot had emerged silks. Moisture content (MO): was measured from the shelled grain for each machine-harvested plot by using a moisture tester mounted on the combine. Grain weight (GW): was recorded in pounds per plot to the nearest two-tenth pound in the field at harvest. Grain yield (GY): The adjusted grain weight was converted from pounds per plot into tons per hectare by using the following formula: (loo-nos) 4.54 GY = GW * * ——————— (loo-15.5%) plot area All the characters described except for days to pollen shed and silking dates were recorded in all locations. These two characters were recorded only at the Central region location. Table 2. 39 Information describing the experimental locations Central (East.Lansing) Northern (Custer’and Stanton) 5211.122: Southern (Cassopolis) Oshtemo Sandy Loam (1989) 6.1 (1990) 6.0 (1991) 6.1 t (1989) very high (1990) very high (1991) very high Temperature ( C°/day) Min. Max. (1989) 10.2 22.2 (1990) 9.8 21.2 (1991) 11.3 23.4 18.41 18.00 (1989) (1990) Rainfall (mm/day) (1989) 614.9 (1990) 932.7 (1991) 808.5 1“ Killing freeze 25‘h Oct. Previous CECE (1989) corn (1990) corn (1991) corn Capac Loam mcnos O O 0 mode; very high-high high-medium medium-low Min. Max. 21.3 21.2 8. 9. 0. 23.0 mmoo 1 .8_Lro a Misting (10‘5 joules/mzlday) 17.75 17.29 578.8 597.4 511.6 19“1 Oct. corn corn soybeans Iosco Sandy Loam, M. M. Sandy Loam 01010) NbUl very high-high very high-medium very high-low Max. 19.4 21.0 22.9 17.75 17.39 425.9 565.7 675.1 259 a 17* Oct. snap beans-rye potatoes potatoes STATISTICAL PROCEDURES ANALYSIS OF VARIANCE A number of missing plots in the tests of the inbred lines and in some F, generations plots were observed in all experiments. Consequently, these entries were deleted from the analysis. A randomized complete block design with three replications was utilized for the analysis of the data, because assumptions related to a lattice design were violated by these missing data. The statistical model used for the analysis of variance in each individual experiment was: Yij=|l+T1+Bj+€1j; Where Y,,= observed value for the in, entry in the j,,, plot; u = overall mean effect : fn= effect of the it entry (i=1,..,68) ; B,= effect of the j“, replication (j=1,..,3) ; and e¢= random error associated with ijm observation. Replications were considered random effects while entries were considered fixed effects. The format of the analysis of variance is shown in.Table 3. The analysis was then combined over locations or years according to the following model: 40 41 Y,j,,=u+T1+Bj+Ek+ (TE) J“38”,; Yfi = observed value of the ijm plot, in the Kt environment ; u = overall mean; *3 ll , the effect of the ith entry (i=1,..,68) 0:: II the effect of the jm replication within the km location or year (j=1,..,3); I; is the effect of the k,I location or year (k=1,..,3); (TE), is the effect of the interaction between the in entry and km location or year; and ‘9 is the error associated with the ijm,observation. Table 3. Format of the analysis of variance for single experiment: Source of Degrees of Mean Expected variation freedom ' squares mean squares Replications (r-1) M3 Entries (e-1) M2 <fl+rKfi Error (r-l) (e-l) M1 02 Total (re-1) ‘ r and e denote the number of replications and entries, respectively. DIALLEL ANALYSIS The method of statistical analysis used is an extension of the analysis given by Griffing (1950) model 1, method 4. 42 Figure 1. The crossing plan of twelve maize inbreds, four um ..1 III” 3-— ._ , JUUL -1 III WEI: Q from each of three maturity classes (early, medium, and late) 1 JLlU ‘Il‘ll'll' ll“ 3 in”HI= 2E Ii- $fifififififififififi “ — indicates maturity classes. indicates the lines within maturity classes. indicates parental lines which were excluded in the analysis. indicates the restrictions which were imposed by the model. 43 This analysis includes partition of general and specific combining ability into among- and within-class effects. The parents were not included in the combining ability analyses. The linear model for a diallel cross among cw lines, w from each of c classes, of F, crosses grown at a single location is: _ I Yijkl'“+g/1j+glk1+s 171:1”:ij Yijk1=(Gi+gij) + (Gk+gkl) + (Sik+sijk1) *erm Where Y&,is.the value of the cross between the jm parent of the it class and the 1, parent of the k1h class; i=1,2,3,...,c; j=1,2,3,...,w; p is the population mean; g/jj=Gj+gjj; CW 2 991:0 1.? g', is the general combining ability effect of the jm parent of the im class; G,is.the part of the general combining ability effect associated with the 1,1, class; 44 C E: G,=0 9,, is the general combining ability effect associated with the j, parent of the 1, class; V s',,,,, is the specific combining ability effect associated with the cross between the j,,I parent of the im class and the 1, parent of the km class; Slijkl=SiK+Sijkfi CV'I I - . 2: S ijkl-o' klvij Si is the part of the specific combining ability effect associated with crosses between parents of the in and k,,, class; C'l kn” S'fi,is.the part of the specific combining ability effect associated with the cross of the jm and 1, parents of the 45 in and k,,I classes; CW1 g: 3111:1‘01' k 911 w(w—1l/2 811111-10; J