UFDC Home  UF Institutional Repository  UF Theses & Dissertations  Internet Archive   Help 
Material Information
Subjects
Notes
Record Information

Table of Contents 
Title Page
Page i Dedication Page ii Acknowledgement Page iii Page iv Table of Contents Page v Page vi List of Tables Page vii Page viii Page ix Page x List of Figures Page xi Page xii Abstract Page xiii Page xiv Chapter 1. Introduction Page 1 Page 2 Chapter 2. Literature review Page 3 Page 4 Page 5 Page 6 Page 7 Page 8 Page 9 Page 10 Page 11 Page 12 Page 13 Page 14 Page 15 Page 16 Page 17 Page 18 Page 19 Page 20 Page 21 Page 22 Page 23 Page 24 Page 25 Page 26 Page 27 Chapter 3. Materials and methods Page 28 Page 29 Page 30 Page 31 Page 32 Page 33 Page 34 Page 35 Page 36 Page 37 Page 38 Page 39 Page 40 Page 41 Page 42 Page 43 Page 44 Page 45 Page 46 Page 47 Chapter 4. Results and discussion Page 48 Page 49 Page 50 Page 51 Page 52 Page 53 Page 54 Page 55 Page 56 Page 57 Page 58 Page 59 Page 60 Page 61 Page 62 Page 63 Page 64 Page 65 Page 66 Page 67 Page 68 Page 69 Page 70 Page 71 Page 72 Page 73 Page 74 Page 75 Page 76 Page 77 Page 78 Page 79 Page 80 Page 81 Page 82 Page 83 Page 84 Page 85 Page 86 Page 87 Page 88 Page 89 Page 90 Page 91 Page 92 Page 93 Page 94 Page 95 Page 96 Page 97 Page 98 Page 99 Page 100 Page 101 Page 102 Page 103 Page 104 Page 105 Page 106 Page 107 Page 108 Page 109 Page 110 Page 111 Page 112 Page 113 Page 114 Page 115 Page 116 Page 117 Page 118 Page 119 Page 120 Page 121 Chapter 5. Summary and conclusions Page 122 Page 123 Page 124 Appendix Page 125 Page 126 Page 127 Page 128 Page 129 Page 130 Page 131 Page 132 Page 133 Page 134 Page 135 Page 136 Page 137 Page 138 Page 139 Page 140 Page 141 Page 142 Page 143 Page 144 Page 145 Page 146 Page 147 Page 148 Page 149 Page 150 Page 151 Page 152 Page 153 Page 154 Page 155 Page 156 Page 157 Page 158 Page 159 Page 160 Literature cited Page 161 Page 162 Page 163 Page 164 Page 165 Page 166 Page 167 Page 168 Page 169 Biographical sketch Page 170 Page 171 Page 172 
Full Text 
RESPONSE OF A TROPICAL LEGUMEGRASS ASSOCIATION TO SYSTEMS OF GRAZING MANAGEMENT AND LEVELS OF PHOSPHORUS FERT ILI ZATION BY RAUL ALONSO SANTILLAN A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1983 Dedicated to my wife, Maggie, my daughters, Alexandra and Carolina, my father and to the memory of my mother ACKNOWLEDGMENTS The author expresses his sincere thanks and appreciation to Dr. Gerald 0. Mott, chairman of the supervisory committee, for his understanding, deep concern and wise guidance throughout the field work and graduate program. Special thanks are given to Dr. W. R. Ocumpaugh, Dr. J. E. Moore, Dr. 0. C. Ruelke, and Dr. L. R. McDowell, who willingly served as members of the committee. Special appreciation is due to Dr. Raul de la Torre, General Coordinator for Animal Production Research at INIAPEcuador, who helped design and advise during the field experiment. The author also expresses his gratitude to Ing. Carlos Cortaza, former Director of E. E. T. Pichilingue and to Ecuador's Instituto Nacional de Investigaciones Agropecuarias for the physical and financial support of this study, especially to the staff of the Programa de Pasto y Ganaderia in the E. E. T. Pichilingue. Outstanding among those whose friendship and physical labors are responsible for this study are Ing. Carlos Cadene, Agr. Jorge Molina, and Senores Cerapio Diaz, Emilio Ortega, Jorge Benitez, Alejo Briones and the late Maximo Soria. Special appreciation is due to Dr. A. I. Khuri, Dr. Michael Conlon and to the graduate student Jose M. Gallo for their outstanding contributions and help in the statistical analyses of the results of this research. Special thanks are offered to both Dr. and Mrs. G. 0. Mott for their generosity to the author and his family throughout their stay in Gainesville. iii The author is indebted to graduate students, Mr. Luis R. Rodriguez and his wife, Teresinha Rodriguez for their friendship and hospitality. Thanks are also due to graduate student Juan Herbs and his wife for their friendship. The author wishes to thank Mrs. Pat French for typing the final copy of this manuscript. Special acknowledgment is due to the author's wife, Maggie, and their two daughters, Alexandra and Carolina, for their invaluable help, love and devotion and finally to my parents and dear sisters, Cecilia and Lucy, for their love and financial help. iv TABLE OF CONTENTS PAGE ACKNOWLEDGMENTS ........................................... iii LIST OF TABLES ............................................ vii LIST OF FIGURES ........................................... xi ABSTRACT .................................................. xiii CHAPTER I INTRODUCTION ................................... 1 CHAPTER II LITERATURE REVIEW .............................. 3 The Tropical Forages .......................... 3 Persistence of Tropical LegumeGrass Associations ......................................... 6 Response to Nutrients ......................... 10 Animal Productivity ........................... 14 Pasture Evaluation ............................ 16 Grazing Systems ............................... 19 Estimates of Dry Matter Production and Yield.. 20 Measuring Botanical Composition ............... 22 Effect of the Grazing Animal on Botanical Composition ................................... 22 Response Surface Methodology .................. 25 CHAPTERIII MATERIALS AND METHODS ......................... 28 LegumeGrass Mixture .......................... 33 Experimental Variables ........................ 33 Experimental Design ........................... 35 Field Plan of the Experiment .................. 35 Land Preparation and Pasture Establishment .... 40 Construction of Physical Facilities ........... 43 Collection of Data in the Three Experimental Years ......................................... 43 Pasture Measurements .......................... 44 CHAPTER IV RESULTS AND DISCUSSION ........................ 48 Effect of Lengths of Rest Period and Levels of Grazing Pressure on Aerial Biomass (DM) ....... 48 Effect of Lengths of Rest Period and Levels Grazing Pressure on the Available Forage (DM). 57 Effect of Lengths of Rest Period and Levels of Grazing Pressure on Grass Yield (DM) .......... 65 v PAGE Effect of Lengths of Rest Period and Levels of Grazing Pressure Upon Legume Yield (DM) ........ 76 Effect of Lengths of Rest Period and Levels of Grazing Pressure on the Yield of Weeds (DM) .... 91 Visual Estimation of Forage Components ......... 103 CHAPTER V SUMMARY AND CONCLUSIONS ........................ 122 APPENDIX .................................................. 126 LITERATURE CITED .......................................... 161 BIOGRAPHICAL SKETCH ....................................... 170 vi LIST OF TABLES TABLE PAGE 1 Soil analysis of experimental site (1978) ........... 34 2 Modified central composite nonrotatable design with four experimental (X) variables, at five levels each, and 41 design points ................................ 36 3 Modified central composite nonrotatable design with four experimental (X) variables, at five levels each, 41 design points, 51 experimental units with their respective area ..................................... 38 4 Aerial biomass production (DM) by year, season, and treatment ........................................... 49 5 Available forage (DM) by year, season, and treatment combination ......................................... 58 6 Grass yields (DM) by year, season, and treatment combination .......................................... 66 7 Legume yields (DM) by year, season, and treatment combination ......................................... 78 8 Yields of weed (DM) by year, season, and treatment combination ......................................... 92 9 Visual estimation of grass percentage by year, season, and treatment combination ........................... 104 10 Visual estimation of legume percentage by year, season, and treatment combination ................... 113 11 Analysis of variance, regression coeffic ents and probabilities for aeiral biomass (g DM/m ) for the wet season of 1978 .................................. 126 12 Analysis of variance, regression coefficients and probabilities for aerial biomass (g DM/m2) for the dry season of 1978 .................................. 127 13 Analysis of variance, regression coefficients and probabilities for aerial biomass (g DM/m2) for the wet season of 1979 .................................. 128 14 Analysis of variance, regression coefficnts and probabilities for aerial biomass (g DM/m ) for the dry season of 1979 .................................. 129 vii TABLE PAGE 15 Analysis of variance, regression coefficients and probabilities for aerial biomass (g 1*1/r2) for the wet season of 1980 ..................................130 16 Analysis of variance, regression coefficients and probabilities for available forage (g DM/m ) for the wet season of 1978 ..................................131 17 Analysis of variance, regression coefficients and probabilities for aerial biomass (g DM/m2) for the dry season of 1978 ..................................132 18 Analysis of variance, regression coefficients and probabilities for available forage (g DM/rn2) for the wet season of 1979 ..................................133 19 Analysis of variance, regression coefficients and probabilities for aerial biomass (g DM/m ) for the dry season of 1979 ..................................134 20 Analysis of variance, regression coefficients and probabilities for available forage (g DM/m2) for the wet season of 1980 ..................................135 21 Analysis of variance, regression coefficients and probabilities for grass yield (g DM/m2) for the wet season of 1978 ......................................136 22 Analysis of variance, regression coefficients and probabilities for grass yield (g DM/m2) for the dry season of 1978 ......................................137 23 Analysis of variance, regression coefficients and probabilities for grass yield (g DM/m2) for the wet season of 1979 ......................................138 24 Analysis of variance, regression coefficients and probabilities for grass yield (g DM/m2) for the dry season of 1979 ......................................139 25 Analysis of variance, regression coefficients and probabilities for grass yield (g DM/m2) for the wet season of 1980 ......................................140 26 Analysis of variance, regression coefficients and probabilities for legume yield (g DM/m2) for the wet season of 1978 ..................................141 viii TABLE PAGE 27 Analysis of variance, regression coefficients and probabilities for legume yield (9 DM/m ) for the dry season of 1978 ................................. 142 28 Analysis of variance, regression coefficients and probabilities for legume yield (g DM/m2) for the wet season of 1979 ................................. 143 29 Analysis of variance, regression coefficients and probabilities for legume yield (g DM/m2) for the dry season of 1979 ................................. 144 30 Analysis of variance, regression coefficients and probabilities for legume yield (g DM/m2) for the wet season of 1980 ................................. 145 31 Analysis of variance, regression coefficients and probabilities for yield of weeds (g DM/m2) for the wet season of 1978 ................................. 146 32 Analysis of variance, regression coefficients and probabilities for yield of weeds (g DM/m2) for the dry season of 1978 ................................. 147 33 Analysis of variance, regression coeffic ents and probabilities for yield of weeds (g DM/m ) for the wet season of 1979 ................................. 148 34 Analysis of variance, regression coefficients and probabilities for yield of weeds (g DM/m2) for the dry season of 1979 ................................. 149 35 Analysis of variance, regression coeffic ents and probabilities for yield of weeds (g DM/m ) for the wet season of 1980 ................................. 150 36 Analysis of variance, regression coefficients and probabilities for visual estimation grass (%) for the wet season of 1978 ............................. 151 37 Analysis of variance, regression coefficients and probabilities for visual estimation grass (%) for the dry season of 1978 ............................. 152 38 Analysis of variance, regression coefficients and probabilities for visual estimation grass (%) for the wet season of 1979 ............................. 153 39 Analysis of variance, regression coefficients and probabilities for visual estimation grass (%) for the dry season of 1979 ............................. 154 ix TABLE PAGE 40 Analysis of variance, regression coefficients and probabilities for visual estimation grass (%) for the wet season of 1980 ............................. 155 41 Analysis of variance, regression coefficients and probabilities for visual estimation legume (%) for the wet season of 1978 ............................. 156 42 Analysis of variance, regression coefficients and probabilities for visual estimation legume (%) for the dry season of 1978 ............................. 157 43 Analysis of variance, regression coefficients and probabilities for visual estimation legume (%) for the wet season of 1979 ............................. 158 44 Analysis of variance, regression coefficients and probabilities for visual estimation legume (%) for the dry season of 1979 ............................. 159 45 Analysis of variance, regression coefficients and probabilities for visual estimation legume (%) for the wet season of 1980 ............................. 160 LIST OF FIGURES FIGURE PAGE 1 Profile of three natural regions of Ecuador, showing the main vegetative zones in relation to rainfall and temperature ..................................... 29 2 Rainfall recorded at Estacion Experimental Pichilingue during the period 19781980 .................. 30 3 Temperature recorded at Estacion Experimental Pichilingue during the period 19781980 .................. 31 4 Solar radiation recorded at Estacion Experimental Pichilingue during the period 19781980 ............. 32 5 Field plan of the experimental pastures ............. 41 6 Effect of rest period and grazing pressure upon grass yield (DM) for the wet season of 1978 ......... 73 7 Contours of grass yield (DM) as affected by length of rest period and levels of grazing pressure in the wet season of 1980 .............................. 74 8 Effect of rest period and grazing pressure upon grass yield (DM) for the wet season of 1980 ............... 75 9 Effect of rest period and grazing pressure upon legume yield (DM) for the wet season of 1978 ........ 84 10 Contours of legume yield (DM) as affected by length of rest period and levels of grazing pressure in the wet season of 1980 .................................. 88 11 Effect of rest period and grazing pressure upon legume yield (DM) for the wet season of 1980 ........ 89 12 Effect of rest period and grazing pressure upon yield of weed (DM) for the wet season of 1978 ....... 98 13 Contours of yield of weed (DM) as affected by length of rest period and levels of grazing pressure in the wet season of 1980 .................................. 100 14 Effect of rest period and grazing pressure upon yield of weed (DM) for the wet season of 1980 ............. 101 15 Effect of rest period and grazing pressure upon grass percentage for the wet season of 1978 ............... 108 xi FIGURE PAGE 16 Contours of grass percentage as affected by length of rest periods and levels of grazing pressure in the wet season of 1980 ............................. 109 17 Effect of rest period and levels of grazing pressure upon grass percentage for the wet season of 1980 ... 110 18 Effect of rest period and grazing pressure upon legume percentage for the wet season of 1978 ....... 119 19 Contours of legume percentage as affected by length of rest period and levels of grazing pressure in the wet season of 1980 ............................. 120 20 Effect of rest period and grazing pressure upon legume percentage for the wet season of 1980 ....... 121 xii Abstract of Dissertation Presented to the Graduate Council of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy RESPONSE OF A TROPICAL LEGUMEGRASS ASSOCIATION TO SYSTEMS OF GRAZING MANAGEMENT AND LEVELS OF PHOSPHORUS FERTILIZATION By RAUL ALONSO SANTILLAN April 1983 Chairman: Dr. G. 0. Mott Major Department: Agronomy A legumegrass pasture composed of glycine [Neonotonia wightii (R. Grah ex Wightii and Amrn.) Lackey], centro (Centrosema pubescens Benth.), guineagrass (Panicum maximum Jacq.), and elephantgrass (Pennisetum purpureum Schumach.) was evaluated in a grazing trial from May 1978 to June 1980 at the Estacion Experimental Tropical Pichilingue, Instituto Nacional de Investigaciones Agropecuarias (INIAP), Quevedo, Ecuador. The main objectives of the study were (a) to determine the effects of length of grazing period (X1), length of rest period (X2), grazing pressure (X3), and levels of P fertilization (X4) upon the pasture mixture; (b) to determine the proper grazing management to attain the optimum legume contribution; and (c) to measure the pasture response in terms of dry matter production and botanical composition. Grazing periods studied were 1, 7, 14, 21, and 28 days; rest periods were 0, 14, 28, 42, and 56 days; grazing pressures were 1.6, 3.3, 5.0, 6.6, and 8.3 kg DM on offer/100 kg body weight; and xiii levels of fertilizer were 0, 100, 200, 300, and 400 kg ha1 of superphosphate. To cover the five levels of the complete factorial (5 4 ), a modified nonrotatable central composite design made up of 41 treatment combinations was used. The response variables included aerial biomass (DM), available forage (DM), grass yield (DM), legume yield (DM), yield of weeds (DM), percentage grass, and percentage legume. A doublesampling procedure was used for estimating pasture production and botanical composition. Rest periods and grazing pressures had the greatest effects on all response variables. Aerial biomass, available forage, grass yield and grass percentage were increased by longer rest periods and by lower grazing pressure. Legume yield and legume percentage were decreased by long rest periods and by low grazing pressure. Short rest periods and high grazing pressures resulted in high yields of weeds. Medium levels of both rest periods and grazing pressure were required for high forage dry matter production and for high legume yield. The other two variables, days grazing and levels of phosphorus fertilization, had negligible effects upon the response of the pasture sward. xiv CHAPTER I INTRODUCTION Ecuador, with an area of 273,670 km 2is located in the northwestern part of the South American continent. The dominant topographical features are two parallel ranges of the lofty Andes mountains that separate the fertile littoral lowland on the west and the more extensive and less fertile lowland of the Amazon Basin on the east. The diversity of natural features of the littoral region is very great due to its multiple climatic conditions, soils, forms of vegetation, and settlement patterns which set it apart from the more homogeneous Sierra and Oriente regions. In the littoral, about 2,500,000 ha are considered as pasture land, supporting 2,875,000 head of cattle giving a carrying capacity of 1.12 animals ha1 The seasonal pattern of rainfall distribution and the low soil N levels are factors restricting forage production and quality during the wetdry seasons. It is well known that cattle production is limited by the feed supply during the dry season, while in the wet season, there is abundance of forage. Pastures are mainly planted to guineagrass (Panicum maximum Jacq.) or elephantgrass (Pennisetum purpureum Schumach.), with other species making up a very small portion of the total hectarage. At present, there is a growing interest in the establishment and utilization of tropical grasslegume mixtures for animal production. There are many advantages to having legume components in the 1 2 pasture. They increase or maintain soil fertility due to their ability to fix N, improve the quality of the diet grazed by the animals, provide better seasonal distribution of the forage throughout the year, especially where alternate wetdry seasons occur. Therefore, a system which combines adequate grazing management practices with highyielding grasses and legumes growing in mixtures which provide feed during the whole year is needed to increase carrying capacity and the final output of the land. The objectives of this research were 1) To determine the response of a tropical legumegrass mixture to various treatment combinations of grazing management factors length of grazing period, length of rest period, grazing pressure, and levels of phosphorus fertilization. 2) To determine the proper management strategy for adapted forage species to attain the optimum legume contribution as a component in the mixture, and 3) To measure the results of legumegrass mixture in terms of aerial biomass, available forage, and botanical composition of the forage on offer. CHAPTER II LITERATURE REVIEW In tropical regions animal production on pastures is dependent upon the quality and quantity of forage available throughout the year. In the humid tropics the seasonal growth of pastures is greatly influenced by the wet and dry periods. At the beginning of the wet season very rapid growth occurs and frequently a large amount of forage accumulates. As the season progresses the rate of growth decreases rapidly and approaches zero during the dry season. The concentration of nitrogen and minerals is relatively high at the beginning of the wet season and also falls rapidly as the season progresses. In addition there is a progressive increase in fiber and decrease in digestibility from the beginning of the wet season through the remainder of the year (Faladines and de Alba, 1963). Some tropical pasture species seem to be well adapted to extreme environmental conditions; however, their potential to supply feed for cattle production may be limited not only by seasonal changes but also by soil fertility (Tergas, 1968), mechanisms of adaptation (Gartner et al., 1974), and grazing management (Mott, 1960; Evans, 1970; Stobbs, 1969). The Tropical Forages Panicum maximum Jacq. has been described by Humphreys (1980) and Bogdan (1977) as a densely tufted perennial grass with relative drought and poor soil tolerance. It is of high nutritive value when young and combines well with other tropical pasture species (Hudgens, 1973; 3 4 Chavez, 1974; Rolando, 1974). It is indigenous to tropical Africa where it is dominant over large areas, particularly under humid and subhumid conditions. In the coastal area of Ecuador it is known as guinea, cauca, saboya, chilena. Betancourt (1969) indicated that guineagrass was the most widespread pasture grass in the lowlands of Ecuador. AcostaSoliz (1967) reported that this grass grows from sea level up to 1400 m. Guineagrass is very popular because of its adaptation to the dry season, resistance to fire, high production of forage, capability of establishment by seed or by division of plant crowns, and finally due to its ability to persist under heavy use and abuse (INIAP, 1980). Recently the Instituto Nacional de Investigaciones Agropecuarias (INIAP) of Ecuador has tested 124 introductions of guineagrass in small grazing trials, mainly for adaptation and persistence. New cultivars will be released to the farmers as soon as sufficient seed is available for large scale use (INIAP, 1979). Pennisetum purpureum Schumach, elephantgrass, is a tallgrowing species, up to 5 m in height. It has been described by Bogdan (1977), Correa (1926), and Mcllroy (1972). AcostaSoliz (1967) and INIAP (1980) stated that elephantgrass grows from sea level up to 2200 m in the warm valleys of the Sierra region. In Ecuador it is mainly used as a pasture grass with some farmers reporting pastures up to 40 years old still under grazing conditions. Under cutting frequencies of 45 days it produces up to 1 l 80 tons DM ha yr when receiving 400 kg of N and irrigation during the dry season (INIAP, 1972). There are three cultivars which are 5 in widespread use in the lowlands of Ecuador. They are 'common' for grazing, 'Hybrid 534' and 'Mexican' for cutting. Centrosema pubescens Benth, centro, a true tropical legume has been described by Humphreys (1980) as a creeping, twining perennial legume native of South America. Grof (1970) reported that the genus Centrosema contains about 70 species growing naturally in the tropical areas of Central and South America. Moore (1962) stated that this tropical legume grows well in humid areas and must be considered as a basic component of pastures under these conditions. Studies conducted at Pichilingue, Ecuador, byHudgens (1973), Chavez (1974), Rolando (1974) and INIAP (1980) indicated that this legume performs well in association with guineagrass and it is also highly persistent under grazing and produces large amounts of seed (Farfan, 1974). In the last few years 132 native ecotypes of Centrosema have been selected and tested in Ecuador. A few of these have been distributed to farmers and are showing some advantages over the Australian commercial cultivars such as higher DM yield, better tolerance to insects and disease, and better adaptation to Ecuador conditions (INIAP, 1979). Neonotonia wightii (R. Grah. ex Wight and Arn.) Lackey, glycine, also known as perennial soybeans, has been described by Humphreys (1980) as a perennial plant, slender, twining, and with long stems having some capacity for rooting at the nodes. Glycine was first introduced into Ecuador in 1966, but it was not used in pastures until 1973 when it was shown to be one of the best legumes for humid and subhumid areas. Three cultivars have been distributed to farmers and these were selected for persistence, 6 adaptability, seed production and disease and insect resistance. These cultivars are 'Malawi' for lower altitudes, 'Cooper' for medium altitudes up to 1200 m and 'Tinaroo' the highest forage yielder grows well from 50 to 1800 m of altitude, producing large amounts of seed from 8001500 m (INIAP, 1979). Persistence of Tropical LegumeGrass Associations Serrao (1976) suggested that the first requirement for successful use of highyielding pasture legumes was their adaptation to local climatic conditions. Secondly, nutrient requirements must be met to insure high yield and maintenance. And finally, they must persist under heavy grazing to secure a long lasting beneficial contribution to the companion grass, to the soil, and to the grazing animal. Gomez (1978) has suggested some other factors which can affect the persistence of tropical legumes when they are growing in association with grasses. Among the most important are (1) environmental factors such as light, temperature and moisture; (2) growth habits of each species growing in the mixture; (3) nodulation ability and capacity for nitrogen fixation; (4) edaphic factors such as pH, nutrient availability, form of supply; (5) frequency and intensity of defoliation by grazing animals; (6) ability to survive during long drought periods; (7) seed production capacity; and (8) pest and disease tolerance. Ludlow and Wilson (1970) reported that tropical grasses achieve up to three times the photosynthetic rate when compared with tropical legumes. This characteristic obviously gives ecological advantages to C4 grasses, affording them the opportunity to grow faster, dominate 7 and even exclude the C3 legumes from the mixture. Tow (1967) showed that green panic (Panicum, maximum var trichoglume) was much more productive at all light intensities and higher root temperatures than glycine when both species were tested under controlled environmental conditions. Roberts (1974) also studied some of the above factors and included some others which are associated with the stability of legumegrass mixtures. These were palatability of the grass and legume, maximum height of the grass, legume ability to grow under the shade projected by the companion vegetation, and the capacity to withstand trampling. He also suggested that continuous grazing helps the legume to compete more effectively with the grass due to more frequent defoliation than under a rotational grazing system. Kretschmer (1974) reported that in general grasses have a better range of adaptation and also a more vigorous growth habit that allows them to compete always at an advantage over most tropical forage legumes. Lack of legume persistence is attributed to the use of unadapted species and cultivars, improper or no fertilization, incorrect rhizobium and overgrazing (t'Mannetje, 1978). Growth habit and leaf morphology of species that compose a mixture are important characteristics which have direct effects upon compatibility and persistence, due especially to light interception ability of each individual species. Santhirasegaram (1976) in the humid tropic of Peru reported that in a wellmanaged guineagrasscentro pasture, the persistence of the legume was due to its viney growth habit enabling it to climb the stems and leaves of this tall and aggressive grass. Thus, the legume can intercept sufficient 8 solar radiation. They also suggested that the ideal type of twining tropical legume should have a strong stoloniferous or rhizomatous growth habit in order to withstand frequent defoliation and heavy damage by grazing animals. Whiteman (1969) pointed out that frequent defoliation results in low yields of twining legumes such as glycine and Siratro, whereas slight defoliation or absence of defoliation results in a higher contribution of the legumes to the total pasture production. In Africa, Draolu and NabusinNapulu (1980) studied the effect of cutting intervals of 3, 6, 12, and 24 weeks and cutting heights of 3.8, 7.5, 15.0, and 30.0 cm in mixed swards of guineagrass and Stylosanthes guianensis. They concluded that DM production was greatly reduced under the lowest and most frequent cutting treatment. The amount of legume in the mixture also decreased as the cutting intervals were increased and the legume was particularly sensitive to close defoliation. In Australia, Mclvor et al. (1981) found the same linear tendency with Desmodium intortum and Setaria sphacelata mixtures under different cutting heights of 8 and 20 cm and cutting intervals of 3, 6, and 9 weeks. The growth response of desmodium was markedly depressed at the lowest cutting height of 8 cm and the shortest cutting interval of 3 weeks. They concluded that cutting at the height of 20 cm at intervals of 6 to 9 weeks was necessary for the persistence of the desmodium in the mixture. Bryan et al. (1971) indicated that the short growth habit of the tropical legumes Stylosanthes humilis and Lotononis bainesil, which may be shaded by taller companion grasses, benefits from heavy grazing pressure which allows light penetration into the canopy. They also 9 concluded that the dominance of climbing legumes such as glycine and Siratro is generally enhanced by light grazing pressure and also by long intervals between grazing periods. Grazing experiments conducted in the wet tropics of Ecuador by Berrezueta (1975), Chavez (1974), INIAP (1979) and Zapata (1981) showed that guineagrasscentro pastures and guineagrassglycine pastures are very persistent and productive mixtures even if heavy grazing pressures are applied. They also observed that rest periods of over 28 days during the wet season favored the companion grass and reduced drastically the amount of legume. Finally they concluded that the guineagrass is dominant, especially at the beginning of the wet season, probably due to the large amount of nitrogen stored in the organic matter during the six to seven months dry season. The high growth rate of this grass decreases slowly and reaches the lowest rate of growth at the end of the dry season while the legumes appeared to be more productive during the dry season. The legumes are the main source of feed for the grazing animals during the dry season. A study of eight grasslegume associations was also made from 1971 through 1973 using grazing animals. Paragrass (Brachiaria mutica)glycine and guineagrasscentro mixtures were shown to be the most compatible and the best accepted by the grazing animals, while tropical kudzu (Pueraria phaseoloides) and Calopogonium mucunoides were unpalatable species, a characteristic that could determine the dominance of this species over the companion grasses (INIAP, 1974). 10 Response to Nutrients Legumes are generally more sensitive to soil factors than other pasture plants, particularly grasses. This sensitivity of legumes emphasizes the importance of understanding the effects of these limitations in tropical conditions. Russell (1978) suggested that improving the growth of legumes in low fertility soils could be approached in two ways: (1) by amelioration of soil conditions through the use of fertilizers or amendments, and (2) by the selection of legume cultivars or genera which are more tolerant to the limiting conditions. There seems to be a general consensus that N and P, in that order, are the plant nutrients that are more often deficient in the tropics (Fox, 1979). The amount of N fixation, nodulation, persistence, and yield of tropical forage legumes may be affected by the soil pH and also by the availability of plant nutrients. Manhaes and Dobereiner (1968)and Fox et al. (1974) reported that for good legume establishment, adequate amounts of available P were required, especially during the nodulation stages. These authors determined that glycine required 60 ppm of P 20 5in the soil solution during the establishment phase. This requirement decreased after the second cutting. In Australia, Andrew and Robins (1969) determined the critical P concentration in the tops associated with maximum plant growth as being the critical levels. These were 0.16 and 0.23% for centro and glycine, respectively. For adequate plant uptake, H 2P0 ions in the soil solution should be between 0.07 to 0.2 ppm. According to Sanchez (1977) some tropical legumes tolerant to low available soil P either absorb P at a faster rate or are able to tranlocate it to the tops more rapidly than do species not tolerant to low P availability (Salinas and Sanchez, 1976; Andrew, 1978). Considerable yield response to P fertilizers was reported by Jones and Freitas (1970) with four tropical legumes (Stylosanthes guianensis, centro, glycine, and Siratro). Similar results were obtained by Franca and Carvalho (1970) in greenhouse studies, using five tropical legumes (glycine var 'Common,' glycine var 'Tinaroo,' Siratro, centro, and Pueraria phaseoloides var 'Javanica' Benth.). In both cases the P deficiency was reflected in decreased nodule weight and N fixation capability. Snyder and Kretschmer (1974) obtained small linear increases in dry matter yields of Siratro, 'Cook' stylo, centro, and Desmodium heterocarpon (L.) DC when lime was applied in 500 kg haI increments up to 3000 kg haI without P fertilization. When the same levels of lime were used together with 45 kg haI of P the response in yield 1 was linear up to 2000 kg ha of lime and curvilinear thereafter. Estimation of P requirements for plant growth should be based on the amount of P needed to give at least 95% of maximum growth (Ozanne and Shaw, 1976). Neme and Lovadini (1967) working with glycine found i that a combination of 120 kg of P205 plus 6 metric tons of lime ha gave a large increase in yield. Werner (1979) reported substantially increased yields of centro to P and K fertilization. Palacios (1976) in Ecuador found that centro responded positively to P fertilization during the establishment period, but yield response was not related to the amount of P applied to the soil during the sowing time. Falade 12 (1975), comparing six levels of P (0, 15, 30, 60, 120, and 180 mg/ 2 kg of soil/pot), found that P concentration in guineagrass and elephantgrass was increased with the addition of P. The same response was reported by VicenteChandler (1975) from Puerto Rico, where 80% of the maximum dry matter production of pearl millet [Pennisetum americanum (L.) K. Schum.] was obtained when soil pH was raised to 5.5 and 115 ppm of P were added. At a pH of 4.8, twice as much P was needed to produce the same forage yield. Fox (1979) reported that the standard P requirement was a relative soil requirement, not an absolute plant requirement. Fox et al. (1974) from Hawaii reported large increases in P uptake by several pasture species, once soils which had high Pfixing capacities were limed to pH 5.0 and 6.0. Phosphorus requirements of soils can range from zero to more than 2220 kg P haI. After N and P, S is considered by many as the next most important element needed for tropical legume growth. Tergas (1977) noted the significance of S on the growth and nodulation of several different tropical forage legumes. Siratro and centro dry matter and nodule weight increased as S was increased. Sanchez (1977) mentioned that S deficiencies are widespread throughout the tropics and that some pasture legumes are more susceptible to S deficiency than most grasses. Medina (1969) reported S deficiencies in some crops growing in the littoral region of Ecuador. Also, a strong response from guineagrass, paragrass, glycine, and centro was observed when S was applied alone or supplied by the ordinary superphosphate or by ammonium sulphate (INIAP, 1980). 13 In the tropics, where there are well defined wet and dry seasons, rainfall plays an important role in the uptake and nutrient content of tropical forage species. Blue and Tergas (1969) reported a drop in N, P, and K contents during the dry season; likewise a decrease of nutrient content during the wet season was found and it was postulated to be due to translocation of nutrients to the roots. Rapid growth during wet seasons may result in trace mienrals being translocated to plant tops where they are rapidly diluted with aerial tissue causing deficiency symptoms in older tissues (Reuter, 1975). Micronutrients can play an important role in tropical legume pastures growth, mainly because of their function in several enzyme systems and in Nfixation by rhizobiumlegume associations. Werner et al. (1975) studied tropical legume response to the micronutrients Mo, Cu, Zn, B, Mn, and Co in the form of FTE BR1O and also in the salt form. Using three tropical legumes planted in pots, they observed visible symptoms of Mn toxicity on glycine and B toxicity on stylo. This work emphasizes that there is a narrow range between deficient and toxic levels of some micronutrients. Medina (1969) found some micronutrient deficiency symptoms in some tropical crops growing in the Quevedo area. He reported that B, Zn, and Fe were the most deficient elements. INIAP (1978) found that at the beginning of the rainy season, the period of most rapid growth for grasses, such as paragrass, Zn deficiency symptoms were evident but these disappear in two or three weeks. Molybdenum also has been recognized as an essential element for legume growth especially for establishment and maintenance. Some authors have suggested that this 14 element is necessary for development of enzymes related to N fixation, nitrate reduction and legume nodulation (Andrew, 1978; Epstein, 1972). Animal Productivity In Pichilingue, Ecuador, Zapata (1981) observed that liveweight gain of steers grazing on common guineagrass, improved Guineagrass Brachiaria humidicola, and improved guineagrassglycine pastures were, l l in the order given, 0.645, 0.678, 0.741, and 0.857 kg animal day1 and an annual liveweight gain of 338.9, 421.9, 458.3, and 540.4 kg ha1 respectively. The same author also noted a difference in liveweight i gain due to the breed of animal, being 0.598, 0.691, and 0.903 kg an1 i day1 for red criollo, braham, and braham x holstein crosses, respectively. Similar results were found at 700 m altitude by Cowan et al. (1974), but in this case milking cows were grazed on guineagrassglycine and kikuygrass (Pennisetum clandestinum) pastures without any i i supplementation. Milk production averaged 9.06 and 12.54 kg cow1 day1 for jersey and holstein cows, respectively. In Australia, Grof and Harding (1970) found that a guineagrasscentro pasture yielded 36% more liveweight than Guineagrass alone, over a two year period. Hall (1970) reported that animal production on unimproved native grass was less than 9 kg ha1 when compared to guineagrassSiratro mixture which i yielded 112 kg liveweight ha . In Costa Rica, Kretschmer (1971) reported that centro in mixtures with guineagrass increased forage yield by 20% during the dry season and by 30% during the wet season, when compared with grass alone. Tergas (1976) reported that centro increased total forage production 15 by 40% when it was grown in mixtures with guineagrass, and the amount l i of N fixed by the legume was estimated to be 100 kg N ha yr In northern Australia, Norman (1970) showed that gain per animal grazing legumegrass mixtures was linearly related to the proportion of legume in the pasture. Also, animal liveweight gain during the dry season was related to the number of days during which the animals had been grazing the legumes. He also reported that animals grazing on grass pastures gained 60 kg head 1, while those allowed to graze grasslegume i pastures gained 280 kg head During the period of 112 days of the l dry season, the first group on grass pastures lost almost 40 kg animal , i while the second gained 60 kg animal due to the companion legume, Stylosanthes humilis. In Ecuador, Chavez (1974) reported that a guineai i grasscentro mixture produced 536.5 kg liveweight gain ha year with the beneficial effect of the legume being more apparent during the dry season, when crude protein content of the guineagrass alone was below 7% and crude protein content for the grasslegume pastures over 10%. Similar results were reported by INIAP (1979) on guineagrassglycine i i var 'Tinaroo' mixture, which produced 458 kg of liveweight ha year In both cases the results were obtained under rotational grazing using the putandtake technique developed by Mott (1960). According to Minson and Milford (1967), tropical legumes maintain adequate nutritive values for a longer period of time than most tropical grasses, when each were under the same management system. The critical level of crude protein required in a pasture before intake is reduced by protein deficiency is estimated to lie between 6.0 and 8.5%. Even highly Nfertilized tropical grasses at late growth stages may have 16 values below these points (Ventura et al., 1975). Tropical forage legumes, on the other hand, retain higher crude protein levels during the dry season, even in advanced maturity stages (Milford and Haydock, 1965). Selectivity by grazing animals is a possible explanation for higher performance of grazing cattle during the dryseason periods in tropical regions, especially when available forage meets or exceeds animal requirements (t'Mannetje, 1974). Pasture Evaluation Mott and Moore (1970) developed a fivephase scheme for forage evaluation. Such a scheme involves quantity and quality determinations. Phase: I. Introduction and breeder's lines, II. Small plot clipping trials, III. Mob grazing experiments, forage response to grazing animals, IV. Animal response, effect of forage on animal output, V. Foragelivestock feeding systems. Forage quality, in vitro organic matter digestibility (IVOMD) is taken into consideration in the first three phases, while phases IV and V involve animal response in nutrient digestibility, performance per animal and production per hectare. Definitions for stocking rate, grazing pressure, and carrying capacity were given by Mott (1960). Stocking rate is defined as the number of animals per unit area of land, the term bearing no relationship to the amount of forage. Grazing pressure is the amount of forage dry matter on offer per animal per day and carrying capacity, also called grazing capacity, is defined as the stocking rate at the 17 optimum grazing pressure. McMeekan (1956) considered the stocking rate as the most powerful factor, influencing the efficiency of pasture conversion to animal products on a per unit area basis. Petersen et al. (1965) developed quantitative expressions which related grazing pressure and carrying capacity with output per animal. Their quantitative theory suggests that animal gain is constant as the stocking rate is increased up to a "critical point," where the grazeable forage is equal to the amount of forage consumed by grazing animals. If the stocking rate is increased beyond the critical point, then the gain per animal decreases, and the animal gain per unit area also decreases. Conway (1965), using three stocking rates, 1.0, 1.75, and 2.5 animals per acre, found that by increasing the stocking rate from 1.0 to 1.75 animals per acre, liveweight gain per animal decreased. However, with 2.5 animals per acre, the liveweight per animal was drastically reduced resulting also in a reduction of liveweight per acre. Jones (1979) in Australia studied the effect of five stocking rates, namely 0.8, 1.3, 1.8, 2.3, and 2.8 an ha1 in combination with three resting periods of 17, 39, and 50 days using a randomized complete block design with two replications. The area allocated to each treatment varied from 0.02 to 0.24 ha, using just one animal per experimental unit. Increasing the stocking rate had the greatest influence upon the persistence of the legume which in this case was Siratro. Also, the resting period became important to the legume productivity under heavy grazing pressure conditions. Results suggest that under heavy grazing pressure a longer rest period would allow the legume to increase or maintain its reserves of nutrients for subsequent growth. Echandi (1956) suggested that carrying capacity 18 should be determined taking into consideration the amount of litter left in the field as residue after each grazing period. Mott (1973) emphasized the use of variable stocking rates in grazing experiments by maintaining the number of animals in equilibrium with the available forage. He noted that the important advantage of the system was that it permits estimation of the carrying capacity of the pasture and the seasonal changes which occur. Evans and Bryan (1973) pointed out the need for more studies to evaluate different grazing pressures in order to measure the yield of the pasture and persistence of tropical legumes. According to Mott (1973) the optimum grazing pressure must be considered as an optimum range instead of a "critical point" and that such an optimum relates only to animal output and may or may not be the optima for plant species in the pasture. Mott and Lucas (1952), Mott (1960), and Matches (1970) described the putandtake technique for grazing trials. They suggest that the stocking rate must be variable in which the grazing pressure is maintained at a constant level and the stocking rate adjusted as the availability of forage changes. These authors also distinguish the terms "testers" for animals which should remain in the pasture throughout the grazing experiment and "grazer," or putandtake animals, as those used to maintain the grazing pressure at the optimum. Mott pointed out that "if the number of animals per unit area is to give an accurate appraisal of carrying capacity, then this unit of measure must not be fixed but be subject to adjustment, so that the number of animals per unit of forage is maintained at an equivalent level for 19 all treatments" (1960, p. 601602). Serrao (1976) suggested that variable or fixed stocking rates can be used in continuous or rotational grazing systems. He concludes that the putandtake system is more appropriate when plant and animal relationships are to be measured. Grazing Systems Continuous Grazing Heady (1970) noted there is much confusion in the definition of grazing systems which are used for describing the day to day provision of livestock feed from a wide variety of sources such as conserved forage and by direct use of pasture by grazing animals. He also defined continuous grazing as a grazing system in which the animals have unrestricted access to any part of the pasture through a grazing period, which can be a season, a year or more. Spending (1965) suggested a pasture under a continuous grazing system could be called correctly grazed when the amount of removed forage by the grazing animals was equal to the amount of forage daily yield. Continuous grazing is the most commonly used system in the tropics, especially on vast areas, far away from the main consumer centers and in many cases areas without any suitable highway system (Chaverra, 1979). Rotational Grazing A rotational grazing system is defined by Heady (1970) and Heath (1978) as a system in which the animals are allowed to graze the pasture for variable periods of time, normally with a heavy stocking rate, during which the pasture is grazed and ungrazed several times during a grazing season or year. 20 The terms grazing period and rest period are commonly used in the rotational grazing system (Heady, 1970). According to Heady (1970) the grazing period is the portion of the grazing time during which grazing takes place and rest period is the time during which the pasture is not grazed. Comparing both continuous and rotational grazing systems on animal production, Stobbs (1969), in Africa, found that animal production was slightly higher for rotational grazing when it was carried out in three paddocks, but it was lower than continuous grazing when it was done in six paddocks. From Australia, Grof and Harding (1970) reported that a mixed Guineagrasscentro pasture with a carrying capacity of 3.5 an hal produced a liveweight gain of 934 kg hal for 2 years and 1075 kg ha 1 for 2 years for continuous and rotational grazing, respectively. From Pichilingue, Ecuador, Paredes (1974) reported that during the rainy season there were no differences in stocking rate between continuous and rotational grazing, but during the dry season continuous grazing under variable stocking rate was superior to rotational grazing with fixed or variable stocking rate. He also found that continuous grazing with variable stocking rates gave the highest average daily gain of 0.457 kg an day Estimates of Dry Matter Production and Yield Estimation of production and yield are major problems in grazing experiments, because of the heterogeneity of the pasture sward and the amount of time required to obtain an adequate sample. One must decide the number of samples, the area to be sampled, and choose an adequate technique for sampling a highly variable population where cover, density, 21 height, weight, and other factors differ from one species to another (Kennedy, 1972). For many years, hand cutting and weighing of aboveground vegetative parts has been the most popular and useful method for estimating forage yield (Heady, 1970) and for estimating grazing pressure (Mott, 1960). Other methods, such as the simple disk meter, described by Bransby (1975), work on the principle of measuring the height of a disk supported by the resistance and compression of the vegetation. Santillan (1976) proved that the simple disk meter was very satisfactory for use in tropical species. Mott (1974) suggested that estimation of total yield or yield of components is based upon the following relationship: yield per unit area = f (density, height). The total yield of an area of vegetation is related to the density and height of individual components. Ground cover and sward height have been used on different types of grassland to estimate dry matter yield. Where pasture vegetation is utilized by grazing animals, the amount of feed present at any one time may be only one of the factors associated with the intake by the grazing animals. Of interest to the pasture scientist is an estimate of that portion of the pasture which is consumed by the grazing animal since they are very selective of plant species and plant parts, which makes for a more complex situation (t'Mannetje, 1978). A double sampling technique is probably one of the best ways to ensure a more precise yield determination. Eyeestimation in combination with a few harvested samples which act as a control on the observer's accuracy is one of the simplest forms of estimating total forage present or annual production potential of a pasture (t'Mannetje, 1978). 22 Measuring Botanical Composition Botanical composition is a very essential measurement, especially in pastures subjected to grazing conditions, because the number of samples and yield of individual species may vary over a wide range depending upon environmental conditions and management factors. T'Mannetie et al. (1976) indicated that botanical composition can be measured in terms of the yield of component species, the number of plants covering the area and also the frequency of occurrence. Determination of botanical composition and sampling techniques are difficult tasks in pasture research. Some methods have been developed to determine botanical composition; the most common are visual estimation and hand separation of harvested material into component species (Gardner, 1972). Visual estimation is a reliable method for studying pasture species, but in some cases it may be difficult to relate dry matter production of each of the component species, especially when growth habit and density differ widely. Tothill and Petersen (1962) indicated that the weight in situ, as well as the estimation of weight of each individual species, and visual estimation are the most useful methods for surveying vegetation of pasture species. Effect of the Grazing Animal on Botanical Composition There are at least three factors that are known to affect the balance of the grasslegume mixtures in the tropics once the pasture has been established. They are stocking rate or grazing pressure frequency of defoliation, and fertilization. 23 The grazing animal has a direct effect on pasture species due mainly to selectivity, deposition of feces and trampling, and an indirect effect on the soil due to the removal of nutrients by the removal of harvested forage. The animal alters the physical and chemical properties such as structure, texture, porosity, water retention, and in some cases, the accumulation of organic matter at certain points due to fecal deposits. All of these changes have a direct consequence, first on the botanical composition and subsequently on the final performance of that grazing animal (Alarcon and Lotero, 1970). Research workers such as Davis (1967) and Wells (1967) agreed that grazing affects the soil cover and botanical composition which results in (1) a reduction of the basal cover, (2) a reduction of height of species, (3) a loss of soil cover and severe erosion, (4) a reduction of root systems, (5) a reduction in the emergence of new shoots, and (6) weed invasion. Bryan and Evans (1973) found that the proportion of legumes Stylosanthes guianensis, Centrosema pubescens, and Pueraria phaseoboides on the dry matter basis was not affected by stocking rates during the first and second year. In the third year a marked reduction took place on the high stocking rate of 6 animals ha 1, in which the proportion of the legumes fell from 22 to 12%, while the companion grass (Panicum maximum) was more affected by the high stocking rate. In this case the reduction was significant since it went from 78 to 65% and to 38% in the first, second and third year, respectively. The other two stocking rates of 2 and 4 animals hal had little effect on 24 the grass and legume during the three experimental years. Similar results were reported from Africa by Stobbs (1969). Cowan et al. (1975), in the Atherton Tableland, Queensland, Australia, found a highly significant correlation coefficient between milk yield haI and the amount of the legume glycine present in the pasture. They studied four stocking rates, 1.3, 1.6, 1.9, and 2.5 cows ha . On the other hand, while stocking rate was increased i the milk production ha also was increased but the percentage of dry matter of legume present in the pasture decreased when the stocking rate increased. Cowan and O'Grady (1976) reported the same trend under similar conditions in another study carried out later at the same location. Some investigations carried out in Cuba have shown that under heavy stocking rates the legumes fail to persist in the mixture. Febles and Padilla (1972), using a stocking rate of 4 cows hal, found that a mixture of guineagrass and the legumes glycine, Siratro, Stylosanthes guianensis, Desmodium intortum, and Desmodium uncinatum, did not persist more than 36 weeks. Funes and Perez (1976), using six animals ha , also found that the three commercial cultivars of glycine (Tinaroo, Cooper, and Clarence) failed to persist under those conditions and that at 36 weeks the proportion of weedy species was drastically increased. Their final conclusion was that it is better to use light stocking rates and that the legumes are useful in those areas in which the grazing management is extensive. On the other hand, there are some results that show a positive effect of stocking rate upon the proportion of legume in the pasture. Vilela (1979), in Brazil, found that 25 the legume percentage was 8.7, 10.5, and 15.6 for the 0.5, 1.0, and 1.5 animals ha 1, respectively, when it was measured with the point quadrat method. Jensen and Schumacher (1970) observed that the percentage of botanical species is not only affected by grazing animals, but also by environmental conditions such as season, precipitation, rain distribution, temperature, and solar radiation. Taking into consideration most of the above factors, Tothill (1978) noted that if the primary objective of the investigation is to obtain an estimate of botanical composition of pastures in assessing animal production, weight of species is the most suitable value to measure. If rainfall interception or photosynthesis are under study, then cover may be the more appropriate parameter. He also mentioned that the importance of this distinction is that number, weight and cover measurements are comparable in time and space, but are independent of the mode of sampling since they are measured directly and expressed in relation to a unit area. Shaw and Bryan (1976) mentioned that the proportion of species on a weight basis is generally the most useful where the main interest is in pasture production and where samples are cut for yield determination, and botanical composition can be determined by handseparating the sample into component species. They also added that this is the most precise and satisfactory method for yield and botanical composition determinations. Response Surface Methodology Littell and Mott (1975) indicated that the purpose of response surface methodology is to estimate the functional relationship,between a response variable such as yield and an experimental variable or 26 control variables, such as rates of P. They also suggested that the range of values determines the experimental region, and the functional relationship is called the response surface. Factorial arrangements of treatments provide good information, but they require greater numbers of experimental units and more physical resources; with the same amount of resources using response surface methodology, it is possible to obtain results from a much greater number of variables and levels within each variable. These designs, such as the rotatable central composite, nonrotatable central composite and San Cristobal, have been used in grazing trials and have proved their usefulness in obtaining valuable data (Maraschin, 1975; Mott, 1977; Serrao, 1976; Villasmil et al., 1975). Maraschin (1975) and Serrao (1976), both using a central composite design, studied the effect of three variables: grazing days (1, 3.5, 7, 10.5, and 14), resting periods (0, 14, 28, 42, and 56 days), and dry matter residue left after grazing (500, 1000, 1500, 2000, and 2500 kg DM ha1 ) upon the botanical composition of Cynodon dactylonDesmodium intortumMacroptilium atropurpureumLotononis bainesiiTrifolium repens mixture. The great advantage of this design is that it used only 24 treatments instead of 125 that the complete factorial would have required for a single replication, to obtain the coefficients needed to evaluate the grazing management systems. Serrao (1976) reported that the most important factors in determining dry matter yield from pasture and also legume percentage maintained in the mixture were grazing pressure and rest periods. He found that the legume content of the mixture was increased with an increase in length of rest period 27 and that heavy grazing pressure and short rest periods almost eliminated the legumes from the pasture. CHAPTER III MATERIALS AND METHODS This research was conducted at Estacion Experimental Tropical Pichilingue, belonging to INIAP and located 7 km from Canton Quevedo, Procincia de los Rios, at 10 06' S Lat. and 790 29' W Long. Altitude at the site is 64 m above sea level. The average minimum and maximum temperatures are 17.3 and 35.6C, respectively, having a mean annual temperature of 24.3C. Annual precipitation is 2152 mm. About 82% of the yearly rainfall occurs during the warmer months from December through June. The months of February and March have the highest rainfall intensity. By contrast, October and November are the driest months, often registering no rainfall. Annual mean relative humidity is 84% and mean sunlight is 846.2 hours per year, 66% of which occurs during the December to May period (Servicio Nacional de Hidrologia y Meterologia del Ecuador, 1980). Holdridge (1967) located Pichilingue in the tropical moist forest zone (Fig. 1). Figures 2, 3, and 4 present the monthly rainfall range, average temperature and average solar radiation for the years 1978, 1979, and 1980. The soils are classified as Torripsamments. Hardy (1960) states that the chemical analysis of this Pichilingue loam reveals a marked deficiency in available P, but an abundance of available K. The soil N status is fair to medium in recently cleared land, but declines rapidly with cultivation. More recent soil analyses at the experimental site revealed that the amount of P is medium, while B, S, Zn and Mo are low (INIAP, 1979). 28 29 a: 2 u 0 0 Ln w 0 0 co rn a) v cD r*j N I I o 0 C) o o 00 E 0 0 oo 8 C) 0 C) to :r n CY C) 0 > C) H (*Ulotlzozl) UJWOO090009) 4j IS36W IVOWCAU 13M A63A ,v LO to (*UlOlt7lool)(UJWOO92OOGZ) > JLS380J'lV:)Idaiiens 13M UGA co GAIIV ) SONVISSV8D 0100 13M to vnovanEWL GAIIV SONVISM9 0100 '80, (S3SSVW) 031Mrino 1 0 0 r? "0 9 3AIJ.VN) S = 'VA MO ca mxmzin, 2 SON'nSSVJ0 HDIH C) 0 m < 0 Z GA LWN) ,4 4 ' SONVISS"D 0'100 ;, o ca 4 W co w OZV8OeV4lHO $4 4J co C: V8(]NnL CnO:) A0 0 ca NOLIV1393A W i a) 14 ONVqSMB 0100 13MA80 4 CC 44 Culoolstl) LUWOOGZOOZI JLS3hlOJ VOidouens JOW S $4 41 0 44 0 W 0 (*UIOZI09) WWO0020091 0 .LS380J 'IVOWO U 13M 44 Cd Cul 09 tpz) (WWOOSI 009) IS380J IVOW081. A80 Cul i7Z ) WLU009 HVNNVAVS AJ L) NIV300 DIJIC)Vd 30 I z 0 13 P4 ca 0 ca 4j w w 00 W r4 co I 00 0 44 u 0 0 rf 4 cc (1) 44 Cl r4 LL (mm) TTUJUT'e'd 31 41 04 0 u 00 0) 10 ca 0 LL C4 C4 C4 C4 44 C4 (o,) ainava9duia, 32 0 *H 44 *i4 i s~c o 0 c, 0 M0 *o I I LL 4 o 4 04 Go a 0 in 0 Cl 0 60 33 Table 1 presents the soil analysis on the experimental site used in this research. LegumeGrass Mixture The legumegrass mixture selected for the experiment consisted of Malawi glycine [Neonotonia wightii (R. Grah x Wightii and Am.) Lackey], commercial centro (Centrosema pubescens Benth.), guineagrass (Panicum maximum Jacq.), and Hybrid 534 elephantgrass (Pennisetum purpureum Schumach.). Guineagrass and elephantgrass were considered because both are very common and useful grass species in the littoral region of Ecuador, and also because of their highyielding capacity, drought tolerance, and ability to grow in mixture with some tropical legumes. Centro is a native legume and may be the most widespread in the lowlands extending from dry tropical forest to wet tropical forest (6002500 mm of rainfall). Glycine had been selected from previous experiments as a very persistent and productive legume. Experimental Variables The experimental variables were (1) Days grazing (X1)l, 7, 14, 21, 28; (2) Days rest (X2)0, 14, 28, 42, 56; (3) Grazing pressure (X3)dry matter on offer per 100 kg body weight (BW); and (4) Phosphorus (P2 0 5) levels (X4). 34 Table 1. Soil analysis of experimental site (1978). Depth of sample: 025 cm pH: 6.30 Percent organic matter: 4.96 Element ppm Phosphorus 62.5 Potassium 598.8 Sulfur 18.4 Zinc 13.2 Manganese 258.5 Copper 20.9 Boron 12.6 35 For this experiment grazing pressure was expressed as the amount of dry matter on offer per 100 kg body weight (BW) and by the residual dry matter in kg ha1 left after each grazing. Each experimental variable was studied at five levels. Therefore, the treatments comprised a factorial type of experiment of four factors, each at five levels (5 4 factorial) (Table 2). Experimental Design Due to the large number of experimental units required to conduct a 5 4 factorial (626 treatment combinations without replications), a response surface design, namely, a modified central composite nonrotatable design was used. The number of design points (treatment combinations) was determined from the following formula: (see Table 2) No. of design points = 2 4 (l) + 2 4 (2) + (2 x 4) + 1 = 41 Certain treatments were replicated twice (central point was replicated thrice) and these are indicated in Table 2. The total number of experimental units was 51. Field Plan of the Experiment In order to estimate the size of experimental units for each treatment, the following formula was used: (see Table 3) S NdR DG where S = size of experimental unit in m 2 N = kg body weight/pasture/day (assumed 300 kg BW), d = number of days pasture is grazed during cycle, 36 41 41 to U I Ln ON M 4 Ln M M ri Ln M M r4 Ln M M 0 0 10 'J' >1 > Cq M .T .C C4 M T .O M T %D C'J Cn T "D o 0 Ln 00 0 o Ln 00 ca u u u u u cc co (D 4 4 4 4 ri r4 rq 4 4 1 ri 4 r4 A C4 r 4 C4 $4 r4 x N ca ri 4 4 to "q 4 0 C) C) 0 CD CD CD CD CD CD C) CD CD 0 0 CD CD C) CD CD CD CD 0 z a) 0000 cc CDC 0 0 CD 0 CD CD CD 0 0 4 > r4 r4 rI r4 r4 r ri rI M M M M M M M M 44 w w T 41 44 w $4 bo go m cn M cn H a,* C C C C C r C 119 ll 1.9 1 9 '19 1 CA N m Q) co CU m m m m m m m m ID 110 110 '.0 r4 r4 rq ri w w 0 0 $4 4 Cn 0 P4 X 41 H ca ca 0 0 0 0 4j ca ',, M C14 T Z C14 C4 T T CA t IT I:T C*ll C4 0 0 1,0 1,0 0 0 ',0 %D 0 m a) t T i r4 T :T r4 r T T r4 r4 T Ln Ln Ln Ln 0 w H ca 0 10 p to 41 r 14 r 4 rl 4 1 00 4 00 rI 00 ri 00 H > ca C4 CN C4 C4 cq CN C4 C14 C14 C\l cn 0 W 0 co M 0 0 c u u co ri rI r4 ri ri 4 1 r4 4 1 4 ri r 4 CN C4 CN4 C 4 Cq ri w m w rn " rq C: Q) Q) > On r4 r4 4 1 rI 4 4 1 rl rq 4 rf 1 r4 C4 C%4 Cq C4 C4 C14 C l C4 X I H > 0 t44 r4 U C14 4 rA 4 ri rA I ri ri r4 4 1 1 l r4 q 4 C14 C4 CIJ C4 CN C*4 C14 H 44 "a 0 41 co rl r4 r4 r4 ri r4 "i r4 i 4 i rA r i i CN C 4 CN CN4 CN C%4 C 4 C w r4 .M ca i cn T Ln o r, co o, o i cA m .T Ln o r 00 m C) 1 CN M T H 4 r4 4 4 1 1 4 q I C4 C11 CJ 37 1 4J 4J 41 41 u 9: 0 %.o T 0 0 %c T 00 Z 0 %o '.0 "o %0 o >% >1 0 0 Ln 00 0 0 Lm 00 C14 Ln 0 r, Ln LM Ln Ln Ln ca u C) 04 r4 4 C14 C*4 r ri r4 rI rj rj rj j Cn w "4 w I N cc H C ri H i to CD 0 0 0 CD 0 0 0 0 0 0 0 CD CD CD CD CD 41 a) Nd C) 0 0 CD 0 CD CD CD 0 c CD CD CD CD 0 CD 4 > eN T It It T ,I T T T C4 C14 C4 Cq C 4 04 11 C14 w Q) IT 44 4 x 1.11 P41 bo w i. 0 CQ w C 1 1 C 9 1 C! C 1 r a) rf rl P4 00 00 00 00 Ul) Ln Ln Ln rq 00 Ln Ln Ln w cn W S w C 4 CD o %0 0 CD o %o %0 00 00 C) % c 00 00 co 00 00 W ca w x Ln Ln LM Ln C14 C14 Ln C4 C'J C4 C14 co p P4 to r. W N 4 r4 00 rl 00 ri 00 ri 00 rj 00 :T T 1 ,1 T T >% ca C4 C14 C14 C14 rl rl rI 1 4 4 ca W ,I C14 CN C14 C4 C14 C4 C4 C14 0 0 0 0 0 C> C4 C,4 C:) X I Cn CJ CN C%4 CN CN C4 C4 C 0 CD CD C*4 CN 0 C) C:) 10 0 C4 C4 CN C%4 C4 C%4 CN C4 CD C C4 4J r. 0 u I I i CN 04 04 . C4 04 0 0 C 0 0 C) CD C4 0i Ln %.D r CO 0", 0 4 C4 M It Ln %,D r CO M C) 4 E z C4 C4 CII C14 Cq M M M M M M M Cv) M M T T 38 w co H E4 > m ca 0 Ln 0 0 Ln 0 0 a 0 Ln 0 0 Ln a 0 0 a 0 0 0 Ln 0 a 0 0 co 44 41 w v p r Ln Ln rl Ln 0 a Ln r Ln Ln r Ln 0 a Ln 0 0 Ln r 0 0 0 0 j4 4J 0 w w w rl r4 C*4 q rq Ti rq C14 rI rq C14 C14 It 14, 4 04 E4 c4 124 92. > &4 41 $4 44 0 CC) Cooccoo occoo Coo CDC cocco 4.J Q) Ln C 0 Ln a 0 0 C Ln C C Ln C 0 0 C C 0 C Ln C a C 0 Aj U*l Ln r Ln C C Ln r" Ln Ln r Ln C 0 Ln C 0 Ln r 0 C Ln C ri CIA r4 r4 q r4 CN rI ri CN 04 T T C14 En 4 rl rI 4 4 "4 "1 "1 4 "1 rA r4 4 rq Pi 4 CIA C14 1 r4 C4 Cq 44 r4 cl rI W r. H Q) w s 0 ca .rq N z 4 H w a) r4 r4o. Hrq w x WQ),N w a) $4 > ,T r4 ri ri r4 r4 ?i rI r4 M M M M Cn M M M "a w 4) x 4 Q) Ln co 41 41 (U ca $4 t3: 4J z = 0 0 H w b lo 14 C N m m M cn M %D ID 1.0 cn M m m %0 110 %0 %.c 110 %D %D cln cn M cn w cn m m m C4 C4 C4 c c c 0 0 w U P4 41 H m 0 Zt C4 C4 It T Cq C4 IT IT C4 C4 ,1 1 C4 C C "0 C 0 .O .O W CO i I T zr rl r4 T T r4 1 ,1 T r4 4 T T VI Ln Ln Ln co 0 cu u u p co F4 rI W co 60 p m 0 q ri ., w w N ri r rH r ri r r4 r ri r rl rl 4 r rI r r..i 4 00 ri 00 ri 00 4 co (U > > ca Ci 04 04 C14 C14 C14 04 C14 a) ca W r4 10 w rI > 44 r4 H 4.4 Q) lc w 0 41 :$ %D tn T i lZ M u) C4 tn W w z r4 C*4 IT cq Ln cq m M m r4 T C14 4 m m m Iq m C*4 IT 'T ca 04 Ln C4 rC14 cn Q) rq 4 ca m T Ln I'D r w (71% rq m ,T Ln M O' C 4 C4 M ,T E z 1 1 r4 4 4 q 4 r4 CN C4 CN CN 39 $4 H E ca ca z 0 0 0 Ln 0 0 0 0 0 0 0 Ln Ln 0 Ln Ln Ln Aj w a, w 0 0 Ln r 0 0 0 0 Ln 0 0 r r Ln r r C14 0 w w w C*4 C14 T 4t rl C 4 CN cn rI 04 E4 :4 1:4 la. 41 44 (n CD c a a C) o CD CD 0 0 CD c C) p CD CD CD CD CD CD tn C) 0 C) C) CD a 0 Ln Ln 0 Ln Ln Ln 0 0 Ln r CD 0 Ln C) Ln CD 0 r r Ln r r r 14 x C14 'T r4 cn 4 cn w cd CCD 14 N CN 04 (N 44 04 00 w pq C: :1 81* w 1.0 ko %10 %D M M m m CD 0 C) CD c en 0 CD 0 M N . co (D cn rq 1 r4 r4 co 00 00 co Ln Ln Ln Un rq CO Ln Ln Ln 41 CD CN C) 0 %D %.o CD 0 ND %D co 00 0 1.0 00 co 00 00 co 0 ca a) x U') U") tn Ln C4 CN a) cq clq . M N 11 m ri w q m W rI W 11, 'T T cr zr T C14 >, CO cq CN i V4 1 1 rq ri r= CO 4 C $4 0 Z u W 0 CD all r CD IT a\ 110 rq Ln 11 r r4 Cq rl CYN cr cc I w Z a, IT rq T T Ln 'T 1 1 CY) 1 0 a7 a4 1: C l cl ca C r4 CN 14 14 cc rq 4J .0 0 0 8 Ln \Z r 00 C\ CD 4 C14 Cn T V) \.C r 00 M 0 rI Eq E4 z C4 C4 C4 CIJ C4 M M M M M M M M M M ZT T + 40 R = kg dry matter offered/kg body weight/day, D = number of days in cycle, and G = growth rate in kg/rn2 /day (assumed .04 kg/rn2 /day). As an example, treatment 14 days of grazing, 28 days of rest, and 5.0 kg DM on offer/iQO kg BW/day is calculated as follows: S=300 x 14 x .05 = 25_ 2 S 42 x0.004 20 This figure was rounded to 1500 mrn to avoid the inconvenience of odd pasture sizes. All pastures were 500 m 2or some multiple (Table 3). In making this calculation it was assumed that one animal weighing approximately 300 kg was used to graze the pasture during the allotted grazing period and designated grazing pressure. It was also assumed that the growth rate was 4 g/m 2/day. In estimating the size (S) of the experimental unit the smallest was set at 500 m 2in order to avoid difficulty in handling the steers in units of smaller size. A total of 7.3 hectares were required for the 51 experimental units in this study (Table 3). These pastures were randomly distributed as shown in Fig. 5. The sizes of the experimental pastures varied from 500 m 2to 4000 m 2; the larger areas were for the continuously grazed treatments (Fig. 5). Land Preparation and Pasture Establishment Land preparation of the experimental area began in May 1977, after the existing vegetation (common guineagrass) was partially eliminated by an application of a grass killer herbicide (Glyfosateo). At the beginning of July the 6yearold guineagrass pasture was plowed under. 41 ca cc 44 CL "0 44 42 In August the complete area was crossdisced in order to eliminate some weeds and obtain a good seedbed. The area was disced again in late September to eliminate the new growth and some weeds. In the same month, 20 small soil samples were taken from the whole area and mixed together for soil analysis; the samples were taken at 25 cm depth (see Table 1). The next step was to prepare the legume seeds. Glycine and centro seeds were scarified using sulfuric acid for a time of 6 and 8 minutes, respectively. After the seeds were washed and shade dried, they were inoculated using an appropriate Rhizobium strain obtained from CIAT (Centro Internacional de Agricultura Tropical). In early October of 1977, a mixture of glycine and centro was sown using a small grain seeder with a four row capacity at the rates of l 3.0 and 6.0 kg ha respectively. The mixture was sown in rows spaced 1.4 m apart. A day after the sowing, a mixture of preemerI i gence herbicides was applied (Linuron 0.75 kg ha + 2 L ha of Laso) to control both grasses and broadleaf weeds coming from seed. Two weeks after the legumes were sown, the grasses were planted vegetatively, using plant divisions for guineagrass and small pieces of stems with two or three buds for elephantgrass. Both grasses were planted alternatively between the rows of the legumes, so that the distance between grasses and legumes was 0.70 m. It was necessary to make one hand weeding using machetes. Four irrigations of 4 mm each were necessary at 12 day intervals. The first irrigation was immediately after the legumes were sown. 43 In early January of 1978, the P fertilizer was applied taking into consideration the respective levels of simple superphosphate kg ha 1 (whose composition was 18.2% P 2 0 5 and 14% elemental S); at the same time a mixture of micronutrients was applied to the whole area. The composition of this mixture was 3 kg Zn chelate, 4 kg Fe chelate, 3 kg CuSO 49 3 kg Borax, and 0.8 kg Molybdenum nitrate ha 1 . Construction of Physical Facilities In February, the total area was surveyed for the purpose of locating the fence lines, and fourstrand wire fences were built. During the second half of March, the entire area was moderately grazed and the remaining vegetation was mowed at 15 cm height. Wooden mineral boxes were built which were used to supply the following formula: 50% of sodium chloride + 50% of mineralized salt containing 7% P as dicalcium phosphate, 0.48% Zinc sulfate, 0.12% manganese carbonate, 0.14% copper sulfate, 0.32% ferrous sulfate, and 0.006% cobalt chloride. Water tanks were provided in each experimental pasture. An area of approximately 10 ha adjacent to the experiment was available as reserve pasture for 50 animals used for adjusting the stocking rates in the experiment. .Collection of Data in the Three Experimental Years. On May 12, 1978, experimental grazing was initiated. Fiftysix Criollo and Holstein/Brahman steers were used to graze the experimental pastures. At the start of the experimental period, the 44 animals were 16 months old and their average weight was 268 kg. Every 28 days all animals were removed from the pastures and from the reserve pasture. They were taken to the main corral for tick treatments and every 56 days they were weighed in order to have a basis for estimating stocking rates for the next grazing cycle. At the end of the first experimental year, the pool of animals was removed from all the pastures and showed average weights of 428 kg. A new pool of younger animals replaced the first group for the second year, having an average weight of 302 kg. The animals for the second experimental year were from the same herds as those of the first year and were 18 months old. Pasture Measurements Dry Matter Determination Before and After Grazing Dry matter ha 1 estimates before (on offer) and after (residual) each grazing period were made in order to apply the required grazing pressure and to determine the net dry matter production. Stocking rate was determined for grazing pressuregrazing period combination on the basis of the total dry matter available before each grazing plus the estimated growth rate during the grazing period. Growth rate during the grazing period was assumed to be the same as that of the previous rest period. The accuracy of this technique for determining stocking rate was checked, using the residual dry matter left on the field after each grazing period. Grazing pressure was also based on visual observations of the amount of dry matter during each grazing period, in order to add or remove animals from pastures 'A 45 depending on the specified dry matter on offer and days of grazing. In order to estimate the grazing pressure, the following formula was used: G/P (A r + gd) S RNd where G/P grazing pressure in terms of 100 kg BW, A available forage before grazing, r residue for a specified grazing pressure, g growth ratepreceding rest period, d number of days for grazing period, 2 S size of the experimental unit in m R kg dry matter offered/kg body weight/day, and N kg body weight/pasture/day. A double sampling technique was employed. As soon as possible, before and after each grazing (during the first 3 1/2 months only), 15 areas measuring 1.0 m 2 were randomly selected. From September 30, area measurements of the same size were taken, corresponding to the circular frame of a forage disk meter (one m 2 ) similar to the one described by Bransby (1975) and used in doublesampling measurements. In each sampling unit the percent dry matter yield of each component of the mixture was visually estimated, followed by an estimate of the dry matter yield in situ. The disk meter was lowered on the forage and after a settling time of approximately 5 to 10 seconds, the disk height (in centimeters) was read off a graduated scale mounted on the center shaft of the disk meter and recorded. During the first three and onehalf months, from the 15 sampling units, three were randomly selected and clipped at ground level for 46 dry matter yield and botanical composition. In September, five sampling units out of 30 disk readings were randomly selected and clipped for the above determinations. These samples were later hand separated into their components, placed inside cloth bags properly identified and dried for 20 hours at 72*C. The sum of the dry weights of the components yielded the total dry weight of the sample. Sickles were used for harvesting the forage samples. The clipped samples were used to adjust through regression analysis the 15 or 30 diskmeter readings of dry matter yield. The forage meter readings and visual estimates of dry weight were used as independent variables in regression equations to generate the regression coefficients needed for calibration of the disk meter and for adjustment of the visual sample estimation. The visual estimate of percent yield was made for the component grasses (guineagrass and elephantgrass), legumes (centro and glycine), and weeds. Although the above grasses and others such as Paspalum fasciculatum, P. paniculatum and Eleusine indica were present in scattered, small patches in some pastures, they were included in the weed components. Likewise, some native Desmodiums were accounted for in the legume component. The response of the pasture mixture to the experimental variables was measured in terms of the following parameters: (1) Aerial biomass (DM) kg ha = grass (DM) + legume (DM) + weeds (DM), (2) Available forage (DM) kg ha = grass (DM) + legume (DM), (3) Grass yield (DMI), 47 (4) Legume yield (DM), (5) Yield of weeds (DM), (6) Grass percentage, and (7) Legume percentage. The above parameters were statistically analyzed for the partial wet season (MayJune) of 1978, the dry season of 1978, wet and dry seasons of 1979, and wet season of 1980. The data were processed using the programs RSREG for response surface design, and 63D for plotting the threedimensional graphs of the Statistical Analysis System of the Northeast Regional Data Center of the University of Florida. CHAPTER IV RESULTS AND DISCUSSION Only two of the four experimental variables included in this experiment, namely, lengths of rest period and levels of grazing pressure are discussed in this section. The other two variables, days grazing and fertilizer level had negligible effects upon the response of the pasture sward. Each of the response variables will be discussed in a separate section beginning in the wet season of 1978 and ending in the wet season of 1980. Effect of Lengths of Rest Periods and Levels of Grazing Pressure on Aerial Biomass (DM) The effect of the lengths of rest period and levels of grazing pressure on the aerial biomass is presented for each of the five seasons in Table 4. The total biomass is the average amount of dry matter present before each grazing period for the rotational grazing treatment combinations, and for each grazing period of 56 days for continuous grazing. Biomass Production (DM) for the Wet Season of 1978 During the first wet season only lengths of rest periods indicated an effect (P < 0.01) on the aerial biomass produced (Appendix Table 11) The linear components of the model accounted for 20% of the total variation, while the quadratic effects and interactions represented only 4 and 13% of the total variation, respectively (Appendix Table 11). 48 49 00 En r %D .1* 0 o w r r N M 11 tn r4 11 00 Cq ND rI CN r r, Ln CN (7N  'T M 0% Ln r ON r a% CD 00 r L'I T %.0 a T Ln 00 00 It 110 r4 :3: cj m m m t c,4 m Ln c,4 r m Ln c 4 r m i i r, m Zo 00000000 coca 00000 0000000 r M tn %Q %0 00 Ln 04 M M r4 0 rI r4 CO 00 00 r r M ON %D Ln Ln M 00 Ln %D 00 zt 00 0 r Ln C4 r tn 0 M CS r .T C4 lZ r 0 M 00 T 04 04 C14 rl ,t C14 M T CsJ M 04 Cq M C4 Ln C4 f4 r4 M C4 ,I M M T m 4+ Co OCCDC 0000 0000 coo 0000000 co P, m 4 M 4 tn 0) M C4 Zr M 04 T 4 M Ln aN M T 1 1,0 1,0 ',C 7 M T M 0 C4 00 T 00 ri Cq 00 M ,T %0 T 00 ,O 4 r CN r r 0 ON M %0 ON :3 60 M M M M %zr M zr tn M T M T C4 04 *,C M I Ln %T M M %.D Ln .hd w Co 1.+ 000000000000000000000000 rz r En r4 M ',D Cn ON Ln 04 %T 4 C14 00 M C4 T C) 0 T C4 C14 O Zr 00 0 Ln 4J T r r Ln %D 0 0 %0 M M 00 C4 1.0 M*Zr r4 C4 1 %0 Cq Ln Ln 00 rCo C*4 C%4 CN C4 M CY) Z T rI C14 04 Cq C14 ,T M rl i M CN M M T Ln w $4 41 Co 0000000000 M m %0 rl 1 IT Ln Ln C) ON Crs V) r Ln Ln M Ln M 00 0 00 1.0 m r C7% %D 00 M r4 r M r4 00 M U* C'J rA 00 0 r M r T C*4 00 M T r.I C r4 rI CN r4 Cq C4 C4 rI M CN C14 rI CIJ CY) i Vn l 4 M 04 Cq 4 M M 00 CD 0 C, C:, 0 C:, a C:, C) 0 C) 0 C, C:, 0 w Ca Cu w w CL Ca Q) rq r4 Cq rI 1 C*q w 04 >1 Ca 11 hd 0 .r.1 41 0 w . . . . . . M M M M M "o "o .o %C M M M m ,o rf '.4 r4 O 0 0 0 X Ca E Cu 0 Co 04 Cq T zt C4 CIJ .T ,t C4 C4 ,T Z C4 CN ,T T C14 CD 0 NC %.0 o 0 o "o 4 r4 31 IT 14 "q T T 14 11 I:r It 14 r4 IT IT Ln tn Ul) Ln Co Q) r4 r 4 r r rA r rI r rq rI 00 4 00 r4 Co rq 00 Cq Cli CN (N C14 C14 Clq C*4 C14 Cq rq rq Q Ca MIT Ln"orW Maicl4mT Ln D rw mc f M 'T z 4 rq r.q r4 ?I rf 1 r4 r( 4 CNJ N" Cq CN 50 0++ 00 rn r4 Cli Ln M w %o IT %D ID C14 m m ol ID ON %0 a ON 0 w Ln 0 %0 r4 Ln rI a,% T w 0 rl Ln r Ln (3) m C14 m IT m It r %D Clq M rl It C*4 m m cq ON ++ 00000000000000000 r4 W m ON Ln I'D tn Ln a m w 0 Ln N .o rq M r4 "  rI 0 r4 Ul) %.0 rq o w r CIN %0 m m C14 fl w IT rq cm N r4 N C14 m m Ln N ri m m m N N N $4 w t44 M m (14 w ON %0 0 r Ln M m Ln W %0 m C14 0 m N N 0 It ri 1.0 N T 0 rq 1i 14 co ++ 00000000000000000 w 04 a% Ln M CN r 00 1 r C14 % O P1 0 Ln rl r "0 1 C4 r 'T %0 %D Ul) 00 00 r T r r a*, a% Ln rl cn 0 ?I gm r4 CIJ C4 T ,T C4 r4 M r4 C14 C4 CN CN a co 44 CD CD CD CD C) 0 C) CD C) CD 0 CD CD 0 0 0 CD w r Ln rn CD Ln CD M r rl M r r CYN M T cn %0 o %o $4 0., a a% m r %0 IT T r Ln Ln %D r4 00 Ln 00 0% cn 0 rq C4 rI Cq rl 04 CN M CN 4 M i rI i Cq to ca r4 r4 C14 C14 ri I C14 04 q q 14 "4 r4 r4 q q M N co r4 I a4 Ic CD C CD CD CD CD CD 0 CD CD CD CD CD CD CD C) C) u rM4 CD CD 0 CD CD CD 0 0 CD 0 0 CD CD 0 CD C) to T T ZT %T T ,t T ,T Cq CIJ C4 CN C4 C4 4T CN 4j 14 m a) 0 co I P4 'D %.0 cn cn rn cn CD 0 CD 0 o M 0 0 CD "a > . . w cn f 1 4 1 00 00 00 CO Ln Ln Ln U") 4 00 Ln Ln Ln En ca W. C,4 0 a %0 .o a C %0 c CO 00 0 "D 00 00 00 00 00 Ln Ln Ln U*) C4 C4 Ln C4 C14 CIJ Cq C4 H N 0 co En 0 q r4 W r4 W rl M 4 W rf W ,T t ,t C4 C14 C14 C14 q 1 r4 1 4 4 Iq CO w CZ cn M Ln 1.0 r 00 m CD .I M T Ln 1 0 r 00 m 0 rq :3 z cli cli cq C14 C4 M m M m Cn cn M cn CY) en 'T ++ 51 The biomass production varied from 1570 kg DM ha I for treatment 17 (1 day grazing, 0 days rest, 1.6 kg DM on offer/100 kg BW, and 0 kg ha1 superphosphate) to 3750 kg DM ha 1 for treatment 19 (1 day grazing, 56 days rest, 1.6 kg DM on offer/100 kg BW, and 0 kg ha1 superphosphate). The aerial biomass production response to lengths of rest periods can only be explained by the management imposed during the first grazing season. All of the experimental pastures were grazed at a medium grazing pressure for 15 days during the latter part of March of 1978, after which they were mowed at about 15 cm above ground level. The wet season of 1978 was represented by a growth period of about 3 months from the first of April until June 30 when the wet season ended. The production data were obtained during this 3 month period beginning on May 12 after only about a month of regrowth. The pastures with short rest periods were included in the first sampling date and for some of these pastures the length of the grazing cycle was only 21 days which means that they had the opportunity to be sampled twice between May 12 and June 30. Pastures with a longer rest period accumulated more dry matter during the growth periods and, thus, showed a higher amount of dry matter produced. The analysis of variance for the wet season of 1978 is presented in Appendix Table 11. Biomass Production (DM) for the Dry Season of 1978 The biomass production during the dry season of 1978 is also presented in Table 4. The biomass production varied from 790 to 4880 kg DM ha 1, corresponding to treatments 26 (28 days grazing, 52 0 days rest period, 1.6 kg DM on offer/100 kg BW, and 400 kg ha 1 of superphosphate) and 32, respectively (28 days grazing, 56 days rest, 8.3 kg DM on offer/100 kg BW, and 400 kg of simple superphosphate). In general, the lowest values of biomass production corresponded to treatment combinations of short rest periods and high levels of grazing pressure, while the highest levels of biomass production corresponded to the treatment combinations with long rest periods and low levels of grazing pressure. The linear components of the model accounted for 74% of the total variation, while the quadratic effects and interactions represented only 1 and 2%, respectively (Appendix Table 12) The experimental variables, length of rest period (X 2 ) and level of grazing pressure (X 3 ) each showed a linear effect upon biomass production (P < 0.01). There was a suggestion that levels of superphosphate (X 4 ) might be having some effect (P < 0.10) but the effects of days grazing (X 1 ) was nil. In all cases individual quadratic effects or interactions were not significant. The lack of interactions between X 2 and X 3 indicated that both variables were behaving independently. The biomass production was increased as the lengths of the rest period were increased and as grazing pressure was reduced. The analysis of variance for the dry season of 1978 is presented in Appendix Table 12. Biomass Production (DM) for the Wet Season of 1979 The biomass production during the wet season of 1979 for each treatment combination is given in Table 4. Biomass production varied 53 from 1210 to 6630 kg DM ha1 corresponding to the treatments 17 [1 day grazing, 0 rest period (continuous grazing), 1.6 kg DM on of fer/loo kg BW, and 0 kg ha1 of superphosphate] and 31, respectively (1 day grazing, 56 days rest, 8.3 kg DM on offer/iQO kg BW, and 400 kg ha1 of superphosphate). During the wet season of 1979 both length of rest period and level of grazing pressure had an effect upon biomass production (P< 0.01), while no quadratic effects and interactions were noted. The linear components of the model accounted for 67% of the total variation, and only 2 and 1% of the total variation was represented by the quadratic and interactions. As is evident, the biomass production is increased as the length of the rest period is increased. The higher grazing pressures also decrease the biomass production and the greatest biomass production was reached when a long rest period and low grazing pressure were imposed. The analysis of variance for the wet season of 1979 is presented in Appendix Table 13. Biomass Production (DM) for the Dry Season of 1979 Biomass production during the dry season of 1979 for each treatment combination is presented in Table 4. Biomass production varied from 1090 to 5980 kg DM ha 1, corresponding to treatments 26 [28 days grazing, 0 days rest (continuous grazing), 1.6 kg DM on of fer/l00 kg BW, and 400 kg ha 1of superphosphate] and 32 (28 days grazing, 56 days rest, 8.3 kg DM on offer/l00 kg BW, and 400 kg ha of superphosphate). Again, the lowest values of biomass on offer ha corresponded to treatment combinations of short rest periods or continuous grazing 54 with the highest grazing pressure, but both experimental variables acted independently from each other. The linear components of the model accounted for 45% of the total variation, while the quadratic effects and interactions represented only 1 and 6% of the total variation, respectively (Appendix Table 14). An examination of the individual linear effects showed that only the experimental variables, days rest (X 2 ) and grazing pressure (X 3) were affecting biomass production (P < 0.01). The number of days grazing (X 1 ) and fertilizer (X 4 ) had no effect upon biomass production. For each 14 days increase in the number of days rest there was a biomass production increase of 230 kg (DM) ha 1, while for each unit (1.6 kg DM) decrease in level of grazing pressure there was an increase of 480 kg ha1 in biomass production. The analysis of variance for the dry season of 1979 is presented in Appendix Table 14. Biomass Production (DM) for the Wet Season of 1980 The lengths of rest period and levels of grazing pressure had a direct effect upon biomass production for the wet season of 1980 (Table 4). There was no interaction between these two experimental variables during this wet season. The linear components of the model accounted for 43% of the total variation, while the quadratic effects and interactions represented only 8 and 1% of the total variation, respectively (Appendix Table 15). The total biomass production varied from 1460 to 7540 kg DM ha 1 for treatments 18 [28 days grazing, 0 days rest (continuous grazing), 1.6 kg DM on offer/100 kg BW, and 0 kg ha 1 of superphosphate and 31 55 (1 day grazing, 56 days rest, 8.3 kg DM on offer/100 kg BW, and 400 kg ha1 of superphosphate). During this particular season, the amount of biomass production was again most affected by the length of rest period and by the levels of grazing pressure. The highest values of biomass production corresponded to the treatment combinations of long rest periods and low grazing pressures. For each 14 days increase in rest period, there was an increase of 550 kg DM ha 1 of biomass production, while for each unit (1.6 kg DM) decrease in grazing pressure there was a 450 unit increase in biomass production. The analysis of variance for the wet season of 1980 is presented in Appendix Table 15. Summary of Biomass Production (DM) In comparison of aerial biomass production (DM) among seasons and years, large differences, especially between wet and dry seasons, are observed. The principal environmental factors involved in these differences between seasons are precipitation, temperature, and solar radiation (see Fig. 2 of Materials and Methods). Every year the rainy season begins in the second half of December, initially with light showers and then increasing in amount and intensity reaching the highest peak of precipitation in February or March. After this time the amount and duration of the showers decreases until it reaches almost zero during the second half of June. About 90% of the total precipitation falls from December to June with the remaining months almost completely dry. The second important factor is temperature which is always higher during the rainy season reaching an average of 26*C during the wet season, while the average during the dry season 56 is 22* C. The third important factor is solar radiation which is also higher during the rainy season, reaching values of 500600 hours of sunlight while from July to November this value averages 350400 hours of sunlight. These three factors acting together are the main determinants of total aerial biomass production. Another very important factor which must be taken into consideration is nitrogen. This element, at the beginning of the wet season, is rapidly mobilized from the soil organic matter, which has accumulated in the soil during the dry season. Sanchez (1977) has indicated that N in the tropics is rapidly utilized by actively growing plants with the occurrence of the first rains at the beginning of the wet season but that the remaining N is slowly released during the rest of the season. These three environmental factors plus the availability of N have a great impact upon the aerial biomass production. The most important experimental variables influencing the aerial biomass production were days of rest (X 2 ) and grazing pressure (X 3). Each variable behaved independently and there was a direct relationship between the number of days rest which permitted the plants a better opportunity to accumulate reserves for more rapid regrowth and development after each grazing period. A similar trend was also found for grazing pressure which was directly related to the amount of forage removed by the grazing animals which in turn was related to the high removal and damage to the growing points and axillary buds. Grazing pressure was related to the rate of recovery and subsequently the final production. Hodgson and Ollerenshaw (1969) mentioned that if grazing pressure is increased while resting 57 periods are decreased, the frequency and severity of defoliation is increased affecting directly the subsequent regrowth and the total dry matter yields. Harris (1978) reported that although the function of reserves, availability of growing points, and uptake characteristics are influenced by the level of stubble biomass, the relationship between stubble biomass and growth rate relates to the amount of photosynthetic tissues. Effect of Lengths of Rest Period and Levels of Grazing Pressure Upon Available Forage (DM) Available forage in this context represents the sum of the grass and legume component but is exclusive of the weeds. The effect of the lengths of the rest period and grazing pressure upon available forage is presented in Table 5 for the five seasons of this experiment. The available forage is the average of the estimated amount of dry matter present before each grazing for the rotational grazing treatment combinations and for 56 days for the continuous treatment. Available Forage for the Wet Season of 1978 During the first wet season, only the length of the rest period had an effect on available forage (P < 0.01). The linear components of the model accounted for only 20% of the total variation, while the quadratic effects and interactions represented 5 and 12%, respectively (Appendix Table 16). The means of total available forage varied from 1420 kg DM ha 1 to 3720 kg DM hal (Table 5) for treatments 34 (28 days grazing, 28 days 58 0++ 00 (n Ln Ln zr N r Ln r N IT Z m IT N N m w Ln %.o r4 N N 11 Ln N m  Ln ri ri M 0 r CN r Ln Ln W M tn M N \0 %0 Ln Ln W P4 W ,T 14 rl :3: ri Cn M M r C*4 M Ln M C4 Ln C14 Cq r M r m 110 IT Ln Ln 0 rI r Ln eq m m r4 r 0%0 w %D r %0 zr 11 N T ON m %0 w N m 0 r .T 0 r IT 0 N ON r 1.0 r I'D r 1.0 a% w T r4 C N C4 rf 4' N M ,I' N M N N M N Ln N cn C14 .? cn M T Ch ++ ca r M M CN Ln %.0 %10 r It r4 0 M rI r IT a en 00 M 00 Ln M Cj C) m , C*4 r4 00 M 1.0 0 C4 00 r4 C4 1.0 T 00 %0 rI %D 00 M r M W 1.0 M r4 1: 00 CN M M M ,T Cn t Ln Cn t Cn T (N CN %0 M 14 Ln I:r m m %.o Ln le 001+ r W rl N %0 m N r r Ln %0 ON 0 %D 00 00 00 C4 1 0 CO M CD C) %0 Cq Ln Ln r rri C4 N N N M N 3' T rI N N N N T M rI r4 M M M T Ln 0 u 41 00++ r, W :r Ln M m m m Ln M 0 m o aN M Ln a% m w 0 w 0 ON 110 m tn w M r4 .o m rl r m tn C14 m r 0 I'D m fl m N w m It r4 rq CN ri Cq N C14 rI M CN C4 rI CN C4 rI Cn r4 rl CY) C14 r4 M Cn 4J 10 rI rl r4 rl i r4 r4 4 r4 ri r4 i rI rA "i rA r4 4 C 4 C14 rl 1 C14 CN 0 w ca r4 $: 1 ca co (D I. Ic 0 C CD 0 0 a 0 0 CD a CD 0 0 a CD CD CD 0 0 0 0 a a 0 w r4 rl ri r4 r4 r4 4 rI M M M M M M M M :3: "D %0 "0 %DD %,o "0 ,c m m C C4 C C4 1 1 1 1 C4 C 14 C4 1 1 1 14 C 0 C 00 co w m 0 41 44 0 Q) r4 W T 4' C4 ,T ,T CIJ ZT C4 CN 4 ca rA r4 IT T r r Ln Ln Ln Ln co W q p .r4 co r4 r r4 r 4 r ri r ri r r4 r r4 r 4 P r4 r4 00 rI M 4 W rq M C14 C*4 cli C14 C14 C14 CN 04 C4 cli ca C4 M ,t Ln ,o r W M CD 1 Cli T Ln ,o r 00 C C) cj cn IF4 z 4 A 4 1 4 C4 C14 C14 C14 59 C ++ 00 to 4 zr 1.0 M M T r4 1,0 ON C4 M O r Ln 0 0 ()N M 0 tn 0 Ln 4 M 0 M J 4 Ln m It Ln 1.0 N 11 m m ++ W 0 .o rq .o m IC Ln Ln 1.0 ra%  I Ln O r4 1 0 W r ,O ',0 W rq r W r 9z N M M M tn r4 4 4 M C4 C4 04 r4 ON ++ r En Ln M %0 as .T M P cn r m Pl r Ln M M Ln 0) ON  ON 0% cq a r4 Ln 00 cn 0 rq r :T M Ln r4 cn a% N "I :2c ,T 3* M M %C D M M rl ,T 4 M M M C4 nI 14 4.4 00 ++ 0 0 C) a 0 a C) 0 0 0 C) C) C) C) a C) 0 m cn M Ln M rq zT w ,o w 0 q Ln M Ln T 11 14 r T %D o IT w w 1.0 r r w r Ln 0 m rI ri C 4 CN C 4 N Z ,T CN CN r4 M r4 CNJ CN4 C 4 44 ca 00 ++ 00000000000000000 a; r En M rI CN 0 CYN V) r4 T %0 CIJ t M %0 %0 o Cq M 4 m  m 00 en r %,D en T Ln zr T 110 r It 00 aN C14 :) rl :3: r4 ri CN 04 1 CIJ C14 C4 C*4 C4 M r4 Cq I C14 M ca co r4 r4 V4 4 rI C14 C4 r4 rI ri r4 4 1 M N cc w r4 ca +.C 000000cDcoccoccoccD 44 00000000000000 CC:) 4.J 60 IT 4r T T T T Cr I' C*4 C4 C14 04 C14 Cq T w w 0 cc cn 0) PQ >1 (n cu coo ,Ccn cc 0 "o >, M rq 14 14 0 0 0 O L U 14 O L L L X E IM 4j w Ca 04 0 CD 110 0 CD ID 10 co 00 C) %C co co 00 00 00 C:  X Ln Ln Ln Ui C4 C14 U) 04 C4 Cq CII r4 tQ 0 .rq 0 w w ca 0 14 00 rI 00 rq 00 r4 00 r.1 00 IT It It It 14, 11 It 41 C4 CN C4 rq rI 4 1 4 1 CC W LA CZ "0 3: if CD cn M Ln 1.0 r 00 O' 0 4 C14 M %:r Ln %D r, 00 (DN CD rl 3: z cq CIJ CIJ cq M M en M M M M M M en ++ 60 rest, 5.0 kg DM on offer/lOG kg BW, and 200 kg ha 1of superphosphate) and 19 (1 day grazing, 56 days rest, 1.6 kg DM on offer/lOG kg BW, and 0 kg hal of superphosphate), respectively. Considering the individual linear effects, days rest (X 2) was the only experimental variable affecting available forage (P < 0.01). The other three experimental variables did not affect the available forage. Individual quadratic effects or interactions were not significant. The lack of interaction indicates days rest is acting independently over available forage, which means that the amount of available forage increases as the length of the rest period increases. For each 14 days of increase in days rest (X 2), there was 190 kg hal increase in the available forage. Available Forage for the Dry Season of 1978 (DM) Available forage varied from 790 kg DM ha to 4880 kg DM ha1 corresponding to treatment combinations 26 [28 days grazing, 0 days rest (continuous grazing), 1.6 kg DM on offer/lO0 kg BW, and 400 kg ha1 of superphosphatel and 32 (28 days grazing, 56 days rest, 8.3 kg DM on off er/lOG kg BW, and 400 kg ha of superphosphate), respectively. During the first dry season, the lowest values of available forage were found in those treatment combinations of short rest periods or continuous grazing with the highest grazing pressure, while the highest values were the longest rest periods and the lowest grazing pressures. The linear components of the model accounted for 74% of the total variation, while the quadratic effects and interactions represented only 1 and 1% of the total variation, respectively (Appendix Table 17). 61 An examination of the individual linear effects reveals that only the experimental variables days rest (X 2 ) and grazing pressure (X 3 ) had any effect upon available forage (P < 0.01). The length of rest period increased the available forage as was also the case with a decrease in grazing pressure. For each increase of 14 days in the rest period there was an increase of 440 kg ha 1 of available forage and for each unit (1.6 kg DM) decrease in grazing pressure there was a 500 kg ha1 increase in available forage. The analysis of variance for the dry season of 1978 is presented in Appendix Table 17. Available Forage for the Wet Season of 1979 (DM) During the wet season of 1979 the available forage (DM) varied from 830 kg DM ha 1 to 6630 kg DM ha1, corresponding to the treatment combinations 17 [1 day grazing, o days rest (continuous grazing), 1.6 kg DM on offer/100 kg BW, and 0 kg ha 1 of superphosphate and 23 (1 day grazing, 56 days rest, 8.3 kg DM on offer/lOO kg BW, and 0 kg ha1 of superphosphate), respectively. The treatment combinations with the shortest rest periods and lowest grazing pressure levels yielded the lowest values of available forage for this season; also it was noted that in this case both variables X 2 and X 3 were acting independently. There was no evidence of any interactions occurring. The linear components of the model accounted for 69% of the total variation, while the quadratic effects and interactions represented only 2 and 1% of the total variation, respectively (Appendix Table 18). Only the linear effects of the experimental variables X 2 and X 3 had 62 an effect upon the available forage (P < 0.01). Increasing the days of rest and decreasing the grazing pressure increased the available forage during the wet season of 1979. For 14 days increase in rest period there was an increase of 750 kg ha1 of available forage and for each unit (1.6 kg DM) decrease in grazing pressure there was a 500 kg ha1 increase in available forage. The analysis of variance for the wet season of 1979 is presented in Appendix Table 18. Available Forage for the Dry Season of 1979 (DM) For the second dry season, the available forage for each treatment combination is presented in Table 5. The available forage varied 1 1 from 400 kg DM ha to 5980 kg DM ha corresponding to the following treatment combinations: 25 [1 day grazing, 0 days rest (continuous grazing), 1.6 kg DM on offer/lOO kg BW, and 400 kg ha 1 of superphosphate] and 15 (7 days grazing, 42 days rest, 6.6 kg DM on offer/100 kg BW, and 300 kg ha I of superphosphate), respectively. Only days rest (X 2 ) and grazing pressure (X 3 ) showed any effect upon available forage (P 0.01). Each of these two experimental variables was acting independently because no interactions were found. The linear components of the model accounted for 54% of the total variation, while the quadratic effects and interactions represented less than 1 and 6% of the total variation, respectively (Appendix Table 19). For each 14 day increase in the rest period, there was an increase of 320 kg ha 1 of available forage and for each unit (1.6 kg DM) decrease in grazing pressure, an increase of 570 kg ha I of available forage was realized. The analysis of variance for the dry season of 1979 is presented in Appendix Table 19. 63 Available Forage for the Wet Season of 1980 (DM) The average amount of available forage (DM) for each treatment combination during the wet season of 1980 is presented in Table 5. The available forage varied from 340 kg DM ha 1 to 7510 kg DM ha 1, corresponding to treatments: 26 [28 days grazing, 0 days rest (continuous grazing), 1.6 kg DM on offer/100 kg BW, and 400 kg ha1 of superphosphate], and 31 (1 day grazing, 56 days rest, 8.3 kg DM on offer/100 kg BW, and 400 kg ha of superphosphate), respectively. Only the two experimental variables days rest (X 2) and grazing pressure (X 3) were found to be affecting the available forage (P< <0.01). The other two variables and the interactions were not significant. The linear components of the model accounted for 57% of the total variation, while the quadratic effects and interactions represented only 3 and 4% of the total variation, respectively (Appendix Table 20). The other experimental variables, X 1and X 4'had no significant effect upon the available forage. As the rest period was increased by 14 days there was an increase of 760 kg ha1 of available forage. On the other hand, each unit (1.6 kg DM) decrease in grazing pressure resulted in an increase of 650 kg ha1 of available forage. The analysis of variance for the wet season of 1980 is presented in Appendix Table 20. Summary of Available Forage (DM) Available forage may be considered a more important response variable than aerial biomass since available forage takes into consideration only the components of the vegetation considered edible 64 by the grazing animal. The available forage was estimated by determining the botanical composition of the total aerial biomass and summing the grass and legume components. Over the five seasons during which this study was conducted the available forage was greatly influenced by the environmental factors which are discussed in the aerial biomass section. The experimental variables, days rest (X 2) and grazing pressure (X 3), each had a highly significant effect upon the available forage. The response of available forage to these two experimental variables was essentially linear over the range which was studied. The longer rest period permitted an accumulation of reserves for pasture regrowth and the lower grazing pressure retained more of the aerial parts of both the grasses and legumes affecting mainly the growing points, axillary buds, and traditional leaf area for the production of photosynthate which stimulated subsequent regrowth. It is significant that these two variables seem to act independently as no interactions between them were found within the range of levels in this experiment. Similarities in responses to the same lengths of rest period and with the same grazing pressures were observed by Maraschin (1975) and Serrao (1976). In each of these two cases interactions were found between these two variables where maximum yields were obtained at high grazing pressures in combination with long rest periods and low grazing pressures with short rest periods. The above studies were conducted with Cynodon dactylon and Desmodium intortum which have quite different growth habits than the grasses and legumes included in this study. Cynodon dactylon is a stoloniferous grass with the 65 growing points close to the soil surface as compared with guineagrass and elephantgrass, each of which has apical meristems and lateral buds located much higher from the crown. The two grasses included in the current study are more sensitive to levels of defoliation and short rest periods because these grasses do not have rhizomes or stolons to store reserves. The legumes included in the current study are climbing and twining tropical legumes which no doubt are more sensitive to close defoliation than many other species. Effect of Lengths of Rest Period and Levels of Grazing Pressure on Grass Yield (DM) The effects of lengths of rest period and levels of grazing pressure on grass yield are presented in Table 6. The grass yield is that available before each grazing cycle and each 56 days for the continuous grazing treatments. Grass Yield (DM) for the Wet Season of 1978 During this first experimental season no significant effects were shown by any of the four variables (Xi. X 29 X 32 X 4 ) (Table 6 and Appendix Table 21). Grass Yield (DM) for the Dry Season of 1978 The grass yield for the dry season of 1978 for each treatment combination is given in Table 6. The grass yield varied from 400 kg DM ha 1 tO 4550 kg DM ha 1 which was found on treatments 26 [28 days grazing, 0 days rest (continuous grazing), 1.6 kg DM on 66 0++ 00 cn lz w w a Ln It r4 a% Ln 0 a CN .o 00 Ln 0 Ln Ln M .T rm 0 m r Ln m r rq Ln r a% 0 r r %0 (n cn rq 00 r r 14, M r4 1: r4 N M M M tn CN C14 M Ln C14 ri ,O Cn r cn Ln IT Ln %0 M++ r rn %0 Ln M r r Ln It 0 N Ln m %0 %.0 N m cq r r.4 r m Ln 0 M .t Ln a% M ** r c,4 M C14 C14 %.C Ln Ln T M %0 Ln 0 Ln r, 4 rl C,4 rA i Cn ri C14 T r4 C14 C14 4 Ln m C14 It cn (n It m ++ r En r 0 m 00 %0 4 r M m m fl rq T m Ln m 110 rI 1.0 rq N m IT m  %0 1,0 Ln 0 r T 0 T U Cn C14 M ,I r r r Ln a% Ln C% M Cn Ln M ,4 1: 00 ri C4 M M M 04 Zr Ln M Cn C?) CIJ ri Ln Cn tn cn cn cn %0 00 ++ 000000000000000000000000 r En Ln cy) ,c c cn T zr r r cn r, Ln r, c,4 oo Zo %c T %o cl :T cD r, Ln m  %0 r4 P4 r4 a% M M r, Cq 4 i r M C) Cf) .T r ,0 0 r %.0 ,O M Ln 14 C j cq CN (N en cn rl eq 4 C14 m C14 M 14 cq C4 m It 0 u 00 ++ 0 CD C C) 0 0 0 0 C 0 0 0 C) 0 a 0 0 0 0 0 0 0 0 0 W r rn m 14, T It rI m 110 It T Ln ON a m w M 0 M %0 r 0 M M r, M Ln M W 4 IT C4 M M 04 M 4* ,O ON r 0 M W ct T 0 0 0 4 rl r4 r4 P4 ri 4 4 r4 rI r4 rI 4 4 rl r4 rl rq rl C4 ri r4 r r4 r4 rq rq rq rq ri r4 rl r4 4 q ca 0 rq cc 44 to r.4 r4 r.I m m m m m m m cn ca 14 :Bc >1 m 8N o cn m cn % o 0 "o cn cn cn cn "o '.0 "o %0 "D z o z cn cn cn cn U . Cn M en cn cn 10 ID %0 %10 cn cn cn cn %0 %0 110 o Iq 14 4 11 co 00 co 00 X "a cu "q co 0 I T C*4 :r zt CN Cq T ,T C*4 T T C14 0 ,C %0 0 0 C .O >1 4 IT T 11 q 1t IT 14 IT IT 4 rq IT T Ln Ln V") Ln 14 r4 r rq r q r r4 r% 4 r ri r 1 4 W V1 m 4 CO rq W C14 C14 C14 C*4 C14 C4 cq C14 C14 C14 CN cc r4 C14 m IT Ln 110 r 00 m C) i C14 CY) It Ln \.c r co m C) 14 C14 M T z 4 4 j r4 rf 4 r4 1 r4 r4 C 4 CN CN4 C 4 67 ++ 00000000000000000 00 En C14 m r (ON 110 w %o r4 Ln 0 T w 0 m w C7% C% N 14 m 0 w 0 It rl C4 T Ln 0 fl 110 %.o 0 C'4 rI 3: m r m r %.0 N rI zr CJ C4 C14 04 ++ 00000000000000000 ri m IT rI r ND 00 C14 r 1.0 00 0 r CY) co 11 Cq M C> T M ,T IA f CN 00 0 M N ,t a% C14 %0 CN CN Cq M Ln C4 r4 r4 M r4 r4 4 r4 I a% ++ ca a 0 0 a 0 0 C) CD 0 0 C) 0 0 0 CD C) 0 r rn 4 ON IT cli ON 00 04 cn 00 00 Ln 00 %.0 M Ln r4 co rm  r 1 00 IT 110 Ln rl Ln C14 cn r cli r %C rl cn rA 1: 00 IT m Clq cli 110 Ln C%4 C11 4 CY) 4 CA C14 C14 C14 S4 W N H rq 00 ++ 0 0 C) 0 0 0 0 C) 0 C) 0 0 0 0 0 a 0 .11 rl W r 0 T Ln It r4 % o 00 cc% It r %0 rA C> C14 00 $4 M T 0 C Ln C14 r4 ON IT rq 00 0 C) r T fl w CN4 4 T .T rl r4 r4 CN r4 C%4 4 ri r4 t" 11 pr4 00 ++ 00000000000000000 a; M 00 q C14 110 C) IT m en 110 0 cn T ON 11 r" r p 4 m tn rI en IT Ln 4 00 Ln M ri M r It C14 :3 q 1 rI rq r4 rq 4 4 ri ri r ri r4 ri 4 "1 0 w ri rI C 4 rq rl N Cq r ri 4 r4 i A r4 r4 CY) N ca w 0 0 CD 0 C) 0 C C) 0 C 0 C) 0 0 0 CD 0 CD 0 CD 0 CD 0 0 0 0 0 a C> CD CD cc T " CN C4 Cq t w w 0 cc 44 ca cn en m 0 0 0 0 %D M 0 C) C) a >1 . . . . . . 4 Cf) 1 11 ri r4 w w w W Ln Ln Ln Ln rA W Ln Ln Ln 10 41 0. it 0 !:) cn (U IFi ^ Q) CO P4 C4 0 0 110 1.0 0 0 o 110 w m 0 1.0 w w w 00 00 0 0 w  >< U*) Ln Ln Ln CN C*4 Ln CN C4 C4 Cq C14 H lz $4 N 0 H E4 0 w 4.J 4 co 0 w w 0 Ii rI 00 r4 00 r4 00 r 00 4 00 %T Z %T T ,T 1 4.J C14 C14 C14 C4 Pf 1 4 4 ri I M Q) IV co Ln \.O r m 0 14 C14 cn IT Ln .o r m 0 4 rn 3: E N C14 m m m m m m m m m cn t T + ++ 68 offer/100 kg BW, and 400 kg ha1 of superphosphate] and 24 (28 days grazing, 56 days rest, 8.3 kg DM on offer/100 kg BW, and 0 kg haI of superphosphate), respectively. Both rest period (X 2 ) and grazing pressure (X 3 ) had an effect upon the grass yield (P < 0.01). The application of superphosphate (X 4 ) also had a significant effect at the 10% level of significance. Each of these variables acted independently for grass yield since no interactions occurred. The linear components of the model accounted for 74% of the total variation, while the quadratic effects and interactions represented only 2 and 1% of the total variation, respectively (Appendix Table 22). An examination of the individual linear effects shows that the experimental variable, days grazing (X 1 ),did not have any effect on the amount of grass produced. The greatest effects were obtained from days rest (X 2 ) and grazing pressure (X 3 ). As the length of rest period increased from zero days (continuous grazing) the amount of grass yield tends to increase, whereas, when the grazing pressure is decreased from 1.6 to 8.3 kg DM on offer/100 kg BW, the grass yield was increased. For each 14 days increase in rest period, there was an increase of 440 kg ha I of grass yield. For each increment of decrease in the grazing pressure, there was an increase of 430 kg ha 1 of grass yield. The analysis of variance is presented in Appendix Table 22. 69 Grass Yield (DM) for the Wet Season of 1979 The grass yield for the wet season of 1979 for all treatment combinations is presented in Table 6. During this season the grass yield varied from 560 kg DM ha1 to 6530 kg DM ha 1 for treatment combinations 17 [1 day grazing, 0 days rest (continuous grazing), 1.6 kg DM on offer/100 kg BW, and 0 kg ha 1 of superphosphate and 23 (1 day grazing, 56 days rest, 8.3 kg DM on offer/100 kg BW, and 0 kg ha 1 of superphosphate), respectively. Grass yield was again influenced by days rest (X 2 ) and grazing pressure (X 3 ) (P < 0.01), while the experimental variables days grazing and fertilizer level were not significant. There were no interactions between experimental variables during this season. The linear components of the model accounted for 72% of the total variation, while the quadratic and interaction effects represented only 2 and 3%, respectively (Appendix Table 23). The grass yield was positively related to increasing lengths of rest period and negatively related to increasing amounts of forage on offer to the grazing animals. For each 14 days increase in the rest period, there was an increase of 800 kg ha I of grass yield; on the other hand, for each unit (1.6 kg DM) decrease in grazing pressure, there was an increase of 430 kg of grass yield. The analysis of variance is presented in Appendix Table 23. Grass Yield (DM) for the Dry Season of 1979 The grass yield for the dry season of 1979 for each treatment combination is presented in Table 6. Grass yield varied from 210 70 kg DM ha1 to 572 kg DM ha for the following treatment combinations: 25 [1 day grazing, 0 days rest (continuous grazing), 1.6 kg DM on of fer/iQO kg BW, and 400 kg ha of superphosphate] and 32 (28 days grazing, 56 days rest, 8.3 kg DM on offer/lOG kg BW, and 400 kg ha1 of superphosphate), respectively. Again the grass yield was increased by increasing the days of rest and decreasing the grazing pressure (P < 0.01). The variables X1and X4were not significant and no interactions were found between the experimental variables. The linear components of the model accounted for 57% of the total variation, while the quadratic and interaction effects represented only 2 and 5% of the total variation, respectively (Appendix Table 24). For each 14 days increase in rest period, there was a corres1 ponding increase of 420 kg ha of grass yield, while each unit (1.6 kg DM) decrease in grazing pressure increased the grass yield by 540 kg ha 1. The analysis of variance is presented in Appendix Table 24. Grass Yield (DM) for the Wet Season of 1980 The grass yield (DM) for the wet season of 1980 for each treatment combination is presented in Table 6. Grass yield varied from 120 kg DM ha to 7150 kg DM ha for treatments 26 [28 days grazing, 0 days rest (continuous grazing), 1.6 kg DM on of fer/l00 kg BW, and 400 kg ha of superphosphate] and 19 (1 day grazing, 56 days rest, 1.6 kg DM on of fer/l00 kg BW, and 0 kg ha of superphosphate), respectively. Grass dry matter yield was again affected by increasing the number of days rest and decreasing the grazing pressure (P < 0.01). 71 The other two experimental variables, X 1 and X V were not significant and no interactions were found between the experimental variables. The linear components of the model accounted for 59% of the total variation, while the quadratic and interaction effects represented only 5 and 3% of the total variation, respectively (Appendix Table 25). An increase of one unit (14 days) of rest increased the grass dry matter yield by 870 kg ha 1, while a decrease in the grazing pressure increased the grass dry matter yield by 610 kg hal. The analysis of variance is presented in Appendix Table 25. Summary of Grass Yield (DM) During the first rainy season (1978), with less than 2 months duration from May 12 to June 30, the experimental variables had no effect upon grass yield. Beginning with the dry season of 1978 through the last wet season of 1980, the grass yield was positively related to increasing lengths of rest period. The two grasses, guineagrass and elephantgrass, used in this experiment are tall growing species with a high capacity for dry matter production and each of them are favored by long rest periods. As the length of the rest period was increased the dominance of each of these grasses was evident, leaving very little space for other species, such as legumes and weeds. The grass yield also increased with time as these two species of grasses became securely established. The mean grass yield for the wet season of 1978 was 1530 kg DM hal' while for the last wet season of 1980, the grass yield was 3630 kg DM hal. It is evident that the amount 72 of grass increased considerably from 1978 to 1980 (Figs. 6, 7, and 8), due mainly to its rapid growth capacity and ability to eliminate the other companion species. VicenteChandler (1975) reported that elephantgrass and guineagrass. are the most productive tropical grasses reaching values of 78 and 45 tons DM ha year 1. INIAP (1980) reported that under a cutting system, the rate of growth for elephantgrass and guineagrass is in the order of 166 kg and 149 kg DM haday ,' respectively. These values corresponded to those obtained in this experiment during the wet season of 1979 but as the year progressed, these values decreased and reached the lowest rate at the end of the dry season of that particular year, being in the order of 32 and 28 kg DM ha1 day1 for elephantgrass and guineagrass, respectively. Grazing pressure expressed as the amount of forage offered per 100 kg BW increased the grass yield as the grazing pressure was decreased. The lowest yields were obtained under the highest grazing pressure and the highest yields were obtained in the treatments with the lowest grazing pressure. If the two experimental variables, rest period and grazing pressure are considered in combination, then the lowest grass yields were obtained under continuous grazing and the highest grazing pressure, while the highest yields were observed on treatment combinations having the longest rest period with the lowest grazing pressure. As the grazing pressure was increased, the amount of plant shoots that were removed by the grazing animal also increased and when this was accompanied by an increase in frequency of defoliation (short rest periods or continuous grazing), the 73 cu 0 cn p w w 41 W 44 w $4 44 0. 0 to C: CD CD 0 I= 0 0 0 a CD cn H 4 co ull C14 N $4 cli r4 00 ca 0 4 bn Ln >1 to co to 0 W ri Cn m $4 00 cz 41 C14 cu co a > 0 ca H 4 4J O En r C) ,a C) CD C) C) ri rq 00 C4 44 44 WH C14 rf 0 0 r4 I >1 ca u 0 co (U w cc bd 44 ca 44 W w 74 cn W 44 M 0 4 44 0 C> to W Ln PO .0 4J 14, CU $4 44 44 0 0 44 (M ca rn C) 0 ri) CY) C) cc U") bo w N co C) C 00 ,A 41 cn >1 0 U) M cn $4 En 0. ca M 4 > M to 44 0 cl ca co C) 4 z 0 0 Lr) 41 H C: 4 00 0 (3) aN m C14 co Q. q Q) Lr) C CN co cz 75 :3: pq w CD $4 (D 4.4 Ia. 44 0 H 0 0 N 41 CD co 0 p 4 C4 co 0 to tn cn en Q) Q) 4 cu 12, 00 bo C14 "4 N $4 cq co to >1 co 0 Im r4 4J Ul) m 00 Q) 0* C C) 0 C) p rA 0 C 0 ,a 0 4 44 C14 0 0 >1 co u 0 W cn ca t" ca co bo 44 0) U) 76 intensity of plant removal was the major determinant of grass yield. This was true for both guineagrass and elephanatgrass and the effect of severe defoliation was to reduce the vigor and the regrowth capacity of these two species. At this point, some bare soil areas were observed and weed invasion occurred. This was very evident under short rest periods when combined with high grazing pressure. The animals had a tendency to selectively graze the grass species during the wet seasons and to selectively graze the legumes during the dry seasons. Similar results have been reported by Humphreys (1978) who observed a selective consumption of Panicum maximum in preference to Stylosanthes guianensis while in the dry season, S. guianensis was wellconsumed. It is well known that tallgrowing grasses are very susceptible to a higher degree of defoliation, especially if these species are subjected to low cutting heights or high grazing pressures. Most of the leaves, growing points, and axillary buds are located much higher than on shortgrowing species. The two tallgrowing species in this study have long internodes with the axillary buds located far apart on the stem so they are very vulnerable to intense defoliation. Effects of Lengths of Rest Period and Levels of Grazing Pressure Upon Legume Yield (DM) Since one of the main objectives of this research was to determine the optimum combinations of the components of grazing management which would favor the legume component of the pasture and permit a higher order of persistence, the legume yield and the changes which occurred during the five seasons are of primary interest. 77 Legume Yield (DM) for the Wet Season of 1978 Legume yields for the wet season of 1978 for each treatment combination are presented in Table 7. Legume yield varied from 410 kg DM ha 1 to 1380 kg DM ha1 for treatments 2 (21 days grazing, 14 days rest, 3.3 kg DM on offer/100 kg BW, and 100 kg ha1 of superphosphate) and 14 (21 days grazing, 14 days rest, 6.6 kg DM on offer/ 100 kg BW, and 300 kg ha1 superphosphate), respectively. Length of rest period was the only factor which appeared to have much effect upon legume yield (P < 0.01). The other three experimental variables, X1. X 30 and X 41' and all of the interactions were not significant. The linear components of the model accounted for 16% of the total variation, while the quadratic and interaction components represented only 4 and 8% of the total variation, respectively (Appendix Table 26). For each unit (14 days) increase in the rest period, there was an increase of only 60 kg of legume yield. The analysis of variance is presented in Appendix Table 26. This lack of response to the experimental variables was expected since the wet season of 1978 was very short, so the experimental variables did not have time to create differences in response. Legume Yield (DM) for the Dry Season of 1978 Legume yield for each treatment combination during the dry season of 1978 is presented in Table 7. The legume yields varied from 380 kg DM ha 1 to 1160 kg DM ha 1 for treatments 28 (28 days grazing, 56 days rest, 1.6 kg DM on offer/100 kg BW, and 400 kg ha 1 of superphosphate) and 24 (28 days grazing, 56 days rest, 8.3 kg DM on offer/ 78 00 r r4 C'4 "0 ON ,I T i r4 Ln r N lzr aN Ul 1.0 T M C*4 cn r4 (n r4 m rq Ln 1 r4 14 ri 0 r r4 r Ln r Ln r r T a's m r IT 0 ON tn cn r en %.D ON r4 cn r r M N rl .o IT M4 N m r4 % o cn cn 0 r En %0 r Ln 00 C) %lo C\l C14 a a T 0 C) r4 Ln m r %0 as as CN a% 00 Ln m Ln in CN N m 10 N m Ln m M Ln 'T w m r4 'T fi I.T D m 0 00 r W N m m rI rl It w r m W Ln r .T IT ON en It CV) It IT Ln Cl) C*4 %D m 1 0 LA Ln zr 1.0 r" 10 00 Ln ID 11 %0 00 a% Ln IT T Ul) Ln 00 m 11 1 co 0 0 0 0 0 a 0 C) CD CD 0 CD a CD C) CD 0 0 0 0 0 0 C 0 f% U3 NT 14 NT 00 cq aN Ln r r 00 r4 0 M 00 rl :T Ln r4 110 Ln r rq Ln Cli aN IT C14 Ln Ln CYN rq %0 C14 f 0 110 00 m 00 C*4 %0 %0 00 00 00 r 4 cn co V4 ri rA rq C; 0 w ca w w r4 rl r4 r4 r4 r4 rI ri r4 rI i r4 "4 r4 ri r4 C14 C14 r4 4 N ca C) 0 0 C) C) 0 C CD 0 CD CD CD 0 CD 0 0 CD 0 0 0 CD CD CD CD CD CD 0 CD CD CD CD CD CD CD 0 0 CD CD CD 0 w rq r4 P4 rl 1 1 rl cn m on m CY) m m m r4 4 0 44 04 M m cn cn cn m Cl) 1.0 % o M m cn cn 1 1; C4 1 1 14 4 14 C C C C 41 S T .T N C,4 .T ,T N CN .T ,t t Z C4 0 0 1 0 1.0 0 0 o 1 0 1 rI .T It 1 4 It T rI "i a T 4 r4 T It Ln Ln Ln Ln ca 4 u 14 r r4 r rI r r4 r 4 r r4 r rf r r4 r rq r4 W r4 W 4 W r4 W C14 eq C14 C14 C%4 C14 C14 C%4 ca 1 Cq M ZI Ln \.C r 00 M C) Cq M ,T Ln D r 00 0'% C 4 N Cl) ,I p ri 4 4 4 1 4 4 4 I C4 N N 79 00 W4 00000000 cnco 0000 C400000r4 c (2' .0 '.00 '.0I 0 Ln TLd at ++ 00000000000000000 ON' I 1 .0 ~4 T rt '.o'.or L, I a++C00000000 000000000 1:4 4 44 0n L o00 r 0000n000000(n00 C1 4nI 00 ++ 00000000000000000 f. (n r 0 U0) ) 0M'0 M 0 cq' TC a% 0 rC r4 Ln0 m m MLn 14 C m I U 1 r4 r 4 l 4 4 r4 4 1 4 C 4rI q r4rfr4r4CV) YN 00000000000000000 00000000000 00 w ca fn ( 1.0 C",cn cn n 0 m . . ..Ua) C44 CC 4 0 C ir44r4 U") 1.0' r0 D IC C 4 )10 r 0 : 3 X C4 C4 C N C'N CNv.~4 en U)mc nm m m Z' j+ 80 100 kg BW, and 0 kg ha1 of superphosphate), respectively. The legume yield was increased as the result of reducing the grazing pressure (P < 0.01). There was an interaction between days grazing (X 1 ) and grazing pressure (X 3 ) but this interaction did not occur again during the remaining three seasons. There appeared to be no linear effect for days grazing (X 1 ), days rest (X 2 ), and fertilizer level (X 4). The linear components of the model accounted for only 31% of the total variation, while the quadratic effects represented only 1% and the interactions 12% (Appendix Table 27). During the dry season of 1978, there was a tendency for the legume yield to increase as the grazing pressures decreased. Legume Yield (DM) for the Wet Season of 1979 The legume yield for the wet season of 1979 for each treatment combination is presented in Table 7. The legume yield during the second wet season varied from 80 kg DM hal to 970 kg DM ha 1 for treatments 23 (1 day grazing, 56 days rest, 8.3 kg DM on offer/100 kg BW, and o kg haI of superphosphate) and 34 (28 days grazing, 28 days rest, 5.0 kg DM on offer/100 kg BW, and 200 kg ha I of superphosphate), respectively. Only the length of rest period (X 2 ) and grazing pressure (X 3) were found to have an effect upon legume yield (P < 0.01). There were no effects of X 1 and X 4 nor any of the interactions between the experimental variables. The linear components of the model accounted for 34% of the total variation, while the quadratic and interaction effects represented 81 only 6 and 7% of the total variation, respectively (Appendix Table 28). During the 1979 wet season, the yield of legumes began to show a pattern associated with the length of the rest period. For each 14 days increase in rest period, there was a decrease of 56 kg ha 1 0 legume yield. Also, during this season, grazing pressure began to show an effect upon legume yield. For each unit (1.6 kg DM) decrease in grazing pressure, there was an increase of 65 kg ha 1 of legume yield. The analysis of variance is presented in Appendix Table 28. Legume Yield (DM) for the Dry Season of 1979 The legume yield for the dry season of 1979 for each treatment combination is presented in Table 7. Legume yield varied from 0 kg DM ha1 to 910 kg DM haI for treatments 36 (14 days grazing, 56 days rest, 5.0 kg DM on offer/100 kg BW, and 200 kg ha1 of superphosphate) and 6 (21 days grazing, 14 days rest, 6.6 kg DM on offer/100 kg BW, and 100 kg ha1 of superphosphate), respectively. Legume yield was affected by length of rest period (X 2 ) (P < 0.01) and by grazing pressure (X 3 ) (P < 0.05). Days grazing and fertilizer levels had no significant effect and there were no interactions among the experimental variables. For length of rest period (X 2 ) there was both a linear and quadratic effect upon legume yield (P < 0.01). The linear and quadratic components of this experimental variable accounted for 43 and 27% of the total variation, while the interactions among the experimental variables accounted for only 3%, respectively (Appendix Table 29). 82 The quadratic effect of length of rest period (X 2 ) strongly suggests that the range of rest periods included in this experiment may have adequately covered the point of maximum legume yield. The highest legume yields were obtained in the region of 14 to 28 days rest and the yields decreased when the rest periods were reduced to zero or increased up to 56 days. The effect of grazing pressure (X 3) appears to be almost linear. For each unit (1.6 kg DM) that grazing pressure was reduced from 1.6 to 8.3 kg DM on offer/100 kg BW, the amount of legume yield increased by 31 kg DM hal. The analysis of variance is presented in Appendix Table 29. Legume Yield (DM) for the Wet Season of 1980 The legume yield for the wet season of 1980 for each species combination is given in Table 7. Legume yields varied from 0 kg DM hal for treatments with the longest rest period (56 days) to 980 kg DM hal for treatment 30 [28 days grazing, 0 rest period (continuous grazing), 3.3 kg DM on offer/100 kg BW, and 400 kg hal of superphosphate]. There was both a linear and quadratic effect of length of rest period (X 2 ) upon legume yield (P < 0.01). There was a linear effect of grazing pressure (P < 0.05). The effects of days grazing and fertilizer level were not significant and neither were any of the interactions of the experimental variables. The linear and quadratic effects accounted for 45 and 24% of the total variation, respectively, while the interactions represented only 5% (Appendix Table 30). 83 The quadratic effect of rest period (X 2 ) gives a better representation of the legume yield which maximizes within the range of 14 and 28 days. As the length of the rest period is either decreased or increased, the legume yield is decreased. As the grazing pressure decreases, the amount of legume yield increased by 37 kg ha 1. The analysis of variance is presented in Appendix Table 30. Observations and Summary of Legume Yield (DM) Some very significant changes occurred in the legume population of these experimental pastures, some of which were recorded and subjected to analysis, whereas others were observations made over the five seasons of this trial. Drastic changes occurred as a result of the treatments imposed from the end of the wet season of 1978 to the end of the wet season of 1980. Some of these observations are recorded here. When the experiment was initiated on May 12 of 1978, the average legume yield for the period until June 30 was about 928 kg ha1. During this initial period, only rest period showed any effect (P < 0.01) upon legume yield. It is difficult to find any satisfactory explanation for the effect of any experimental variable over this short period of time, but differences in the growth rate of grasses and legumes or the shortterm effects of insects like red spider on centre could be a partial explanation (Fig. 9). During the first dry season (July to December, 1978), the legume yield declined from 640 kg ha I which indicated a linear response to 84 00 On~i CDC)e 1H 4 1 0 04 1 c~bc 0 '0 Lf~ w w cn ci) 4 0)) 0)~~U 00 ." r4>, (') 0L4 Q )0 C C 4 0 co 85 grazing pressure resulting in various degrees of defoliation, selectivity, and effect of trampling. Defoliation and selectivity were probably the main determinants on the amount of both legumes. A small difference is also recorded in the behavior of these two species in that centre appeared to be more tolerant of heavy grazing pressure and the area covered by glycine increased as the grazing pressure decreased. During the dry season of 1979, it was also observed that the grazing animals were selectively consuming more legumes than during the wet season, probably because of the lower quality and reduced quantity of grass for that period. Humphreys (1978) and Norman (1970) observed similar preferences of animals grazing tropical grasslegume pastures. During the wet season of 1979 (JanuaryJune), the average legume yield was 473 kg ha1 which was considerably lower when compared with the predecing dry season (1978) which can be explained by the much greater rate of growth by both of these companion grasses and by their stronger competitive ability which shaded the legumes, especially in the treatment with long rest periods. During the 1979 wet season, some environmental factors such as humidity, temperature, light, and N played a very important role in the high growth rates of the tropical grasses. The dominance of the C 4 grass species began to manifest itself over the C 3 tropical legumes. Mott stated that "physiological differences between tropical grasses and legumes have important implications for legumegrass associations. Since their optima for light, temperature, and moisture differ, it is much more difficult to select compatible grasses and legumes in the tropics than among temperate species where the responses to 86 environmental factors are similar." (1981, p. 3541). He also added that the viney growth habit of several genera of tropical legumes (Calopogonium, Centrosema, some species of Desmodium, Neonotonia, Macroptilium, and Pueraria); they confer an advantage over C 4 tropical grasses in that they are able to climb to the top of the canopy. Devising defoliation strategies that will maintain the regrowth potential of viney legumes is very important. The advantage of short rest periods during the wet seasons in terms of legume population was evident, giving these species more opportunity to survive, persist, bloom, and produce some seed, even at heavy grazing pressures. Under high levels of defoliation, the competitive and shading effects from both of the grasses was greatly limited. Each of the tropical legumes bloomed profusely at the beginning of the dry season (JulyAugust). During the dry season of 1979 (JulyDecember), the mean legume yield was 397 kg ha 1. During this second dry season a few pastures with the longest rest periods were almost 100% grass with no legumes nor weeds being observed. The decrease in legume yield may be partly explained by the excessive compititive ability of the grasses and also by the selective pressure on the legumes during this season. Pastures subjected to continuous grazing or short rest periods showed a greater proportion of legume dry matter in the total amount of available forage. Both legume species, centro and glycine, survived the effect of frequency and severity of defoliation when short rest periods were combined with high grazing pressures. 