It is suggested that, since the influence of soil physical and micro-meteorological factors is known to be of great significance in determining the biological response of plants, the time is now appropriate to restrict the number of conventional field experiments and replace some of them with completely instrumented experiments. Until complete descriptions of experiments are available, the quantitative importance of environment and its interactions with fertilizer and cultivation practices cannot be determined. With such information, a limited number of field trials would yield more applicable information than the large number of uninstrumented trials now in existence. The argument is extended to (1) the limited applicability of pot and glasshouse trials; (2) the variation between replicates in field trials as an aid to soil science; (3) the limited utility of soil surveys; (4) the poor correlation between arbitrary soil extracts and field responses; (5) the misuse of a common test species for assessing comparative soil fertility in several regions; (6) the poor correlation between plant analyses and soil deficiencies
Original Title
THE DOUBTFUL UTILITY OF PRESENT-DAY FIELD EXPERIMENTATION AND OTHER DETERMINATIONS INVOLVING SOIL-PLANT INTERACTIONS
It is suggested that, since the influence of soil physical and micro-meteorological factors is known to be of great significance in determining the biological response of plants, the time is now appropriate to restrict the number of conventional field experiments and replace some of them with completely instrumented experiments. Until complete descriptions of experiments are available, the quantitative importance of environment and its interactions with fertilizer and cultivation practices cannot be determined. With such information, a limited number of field trials would yield more applicable information than the large number of uninstrumented trials now in existence. The argument is extended to (1) the limited applicability of pot and glasshouse trials; (2) the variation between replicates in field trials as an aid to soil science; (3) the limited utility of soil surveys; (4) the poor correlation between arbitrary soil extracts and field responses; (5) the misuse of a common test species for assessing comparative soil fertility in several regions; (6) the poor correlation between plant analyses and soil deficiencies
It is suggested that, since the influence of soil physical and micro-meteorological factors is known to be of great significance in determining the biological response of plants, the time is now appropriate to restrict the number of conventional field experiments and replace some of them with completely instrumented experiments. Until complete descriptions of experiments are available, the quantitative importance of environment and its interactions with fertilizer and cultivation practices cannot be determined. With such information, a limited number of field trials would yield more applicable information than the large number of uninstrumented trials now in existence. The argument is extended to (1) the limited applicability of pot and glasshouse trials; (2) the variation between replicates in field trials as an aid to soil science; (3) the limited utility of soil surveys; (4) the poor correlation between arbitrary soil extracts and field responses; (5) the misuse of a common test species for assessing comparative soil fertility in several regions; (6) the poor correlation between plant analyses and soil deficiencies
Reprinted/rom SOILS AND FERTILIZERS, Vol. XXIII, I960, 307-310.
THE DOUBTFUL UTILITY OF PRESENT-DAY FIELD EXPERI-
MENTATION AND OTHER DETERMINATIONS INVOLVING SOIL-PLANT INTERACTIONS by N. COLLIS-GEORGE AND B. G. DAVEY (School of Agriculture, University of Sydney, N.S.W.) SUMMARY It is suggested, that since the influence of soil physical and factors is known to be of great signi- ficance in determining the biological response of plants, the time is now appropriate to restrict the number of conven- tional field experiments and replace some of them with completely instrumented expenments. Until complete descriptions of experiments arc available, the quantitative importance of environment and its interaction with fertilizer and cultivation practices cannot be determined. With such information a limited number of field trials would yield more upplicuble information than the !urge number of uninstru- mented trials now in existence. The argument is extended to (I) the limited applicability of {lOt and glasshouse triuls; (2) the variation between replicates in field trials as an uid to soil science; (3) the limited utilization of soil surveys; (4) the poor correlation between arbitrary soil extracts und field responses; (5) the misuse of a common test species for assessing comparative soil fertility in several regions; (6) the poor correlation between plant analyses and soil deficiencies. The era of field experimentation, which began in 1834 when J. B. Boussingault set up the first field experiments at Bechelbonn, Alsace, was placed on a modern scientific basis by Liebig's report of 1840. The first field experiments in the form used to-day were established by Lawes and Gilbert at Rotham- sted in 1843. Since then the field experimentalist has sought for and has confirmed the importance of the essential elements in influencing the production of crops in the field. However, a great deal of the evidence for the necessity of specific nutrient elements has arisen from investigations in the laboratory and not from field experiments (even Liebig's startling discoveries of the importance of N P and Kin the field were based on strong circum- stantial evidence available prior to his experimen- tation). The application of the results of field trials led to large increases in agricultural production, and the conversion of formerly unproductive or marginal areas into useful agricultural land. and have generaliy made the difference between negligible and economic production. By comparison, the results of most field experi- ments do not fall into such a category. Results of field trials vary from year to year, and although a statistical analysis may show a significant trend, as a consequence of one or more treatments, it does not give an invariable quantitative response to any one treatment. Often field experiments can show a qualitative difference in type of response from season to season. It has become common practice to persist with a trial for a number of years in order to obtain a statistical "'average" result in an attempt to overcome the unavoidable seasonal variability associated with field work. Statistical evidence shows that from 20 to 25 %. in the best circumstances, of the total variation is associated with experimental error. The most modern techniques of experimentation have not reduced this error term. All experimentalists are aware of the importance of season in determining their results, yet it is uncommon for them to define the meteorological environment, even by such modest means as the Stevenson screen and rain gauge, during the pro- gress of the trial. The transference of routine meteorological data from a central station and their application to experimental sites are unlikely to be satisfactory even when the aspect and situations are similar. Yet, from a survey of the literature or by simple experiment, it can be quickly ascertained that moisture content of the soil, soil temperature and the micro-meteorological factors greatly in- fluence the behaviour of a plant. The uptake of phosphorus, for example, is certainly dependent on soil-moisture content, root temperature and solar radiation, and probably also on the relative- humidity gradient at the leaf surface. The yield of cotton has been shown to be largely a function of It is significant that all of the biological responses the quantity and availability of soil moisture, and to fertilizer or cultural ameliorations which have in the production of pastures the response to been adopted into agricultural practice have been nitrogen fertilizers is determined by air temperature as a consequence of very marked field-trial responses and solar radiation. 307 The number of physical variables (soil and micro- meteorological) known to control biological sponses seems to be expanding, whilst the total number of chemical nutrients seems now to have been reached. The time is perhaps opportune to investigate the reasons for the variations encoun- tered in soil-plant trials, and since methods are now available for measuring many physical soil and meteorological parameters it should be possible either to eliminate part of the error in field trials or eliminate those soil and climatic parameters inconse- quential to plant performance. The number of biological parameters which have been measured in field trials has been in its turn limited, and comparison of grain and straw re- sponses for wheat show that it is very dangerous to translate results from one wheat crop to another even if the soil and are constant. Responses to applied nutrients undoubtedly vary with the parameter measured, for example the response to nitrogen in lettuce is more marked on a wet-weight than on a dry-weight basis. Subtler differences occur when parameters involving con- sumer quality (and hence economic returns) are considered. In the long term, to obtain maximum useful information from a field trial, it is not suffi- cient to limit the measured biological parameters to field weight or even to weight. On the basis of the preceding two paragraphs, it could be said that if more environmental para- meters and more biological parameters were recorded, fewer tria1s could be carried out with the available resources. It is one purpose of this paper to suggest that it is now time to stop the majority of the "unrecorded" field trials. The average trial, when analysed, yields the following type of information: "On site A, on soil B, with biological variety C, using biological response parameter D, from the years E to F there was a statistical average response G H for a certain treatment." The definitions of A and B are not necessarily unambiguous, and the economist may complain that there are not enough treatments to allow him to produce a production curve! The information yielded by this trial could only be rigorously applied to some other site where the sequence of climate and crops had been identical and where the soil parameter of consequence to the crop are the same. The chances of such coincidence and hence of quantitative applicability are remote. Unless there is a deficiency, which is so marked that it virtually prohibits growth, for example, trace elements in the 90-Mile Desert of South Australia, or the need for a basic dressing of phosphate on most Australian soils, there hardly seems any need for the infonnation given by most trials since it does not allow logical quantitative conclusions to. be drawn about future trial responses. As things stand, when once one trial in an area on one soil bas been completed, it either shows a deficiency so marked that no land user can ignore it, regardless of seasonal changes, or it shows marginal responses which the trial organizers cannot be sure will repeated qualitatively, but do know are unlikely to be repeated quantitatively in, say, the next ten years. In either case, there would seem to be no need for successive trials: in the first case, the answer is complete; in the second case it is uninterpretable since no one as yet can quantitatively explain annual variations when "ail" the infonnation is before him and the series of annual trials concluded, These remarks are not so stringently applicable to those trials whose object is to find the cumulative or residual, effects of treatments on one site. Buf even in these circwnstances there seems to be only a weak case for persisting with the trial when the magnitude of the variation bet\veen treatments is of the same order as the variation in any one treatment due to seasonal variables. The time has come to conduct far fewer trials involving soil-plant interactions, and, in particular, almost to eliminate the large number of trials, which in some cases both in their design and in the methods have been repeated annually on one site and are budgeted for almost indefinitely. In the history of fertilizer and soil-plant inter- action trials each major discovery has caused a very careful scrutiny of the results and methods used in the earlier work. For instance, the discovery of the importance of the rhizobia] strain and molybdenum in nitrogen fixation, and then later of the effect of neutralized superphosphate in obtaining effective nodulation in the field on acid soils, led to more careful control in new field experiments, since the older trials were uncontrolled in these respects. A further purpose of this paper is to claim that there is now sufficient evidence concerning soil-plant interactions as affected by environmental factors of soil and to suppose that Jack of recorded information is preventing the interpretation of the results obtained in most "uncontrolled .. field trials, and that a new style of experimentation is needed. If the premise of the importance of the environ- mental factors in soil-plant interactions is accepted, several subsidiary but important conclusions can be developed concerning other aspects of soil-plant relationships. (Environment is used in its widest sense to include all soil physical and micro-meteoro- logical factors which can influence biological response.) (I) The limited applicability of pot and glasshouse trials. Glasshouse trials, using pots and culture *-Botanists, in general, have long accepted the consequences of environment on biological response in their thinking, but only recently have had the techniques to estimate these. Many agricultural experimentalists and some biometridans seem to have ignored or minimized these consequences. 308 are used as of soil fertility. The results are mterpretable m the case of deficiencies which are almost growth-restricting. The application of glasshouse results in toto, bO\Vever, is a hazardous procedure. Our hypothesis leads to the conclusion that the general transference of such results to field behaviour would be possible if the relationship of biological responses to environ- ment were known, and the environments of the glasshouse and tbe field situation had been deter- ruined quantitatively. It is unlikely that the trans- lation will be accomplished for some time. To transfer evidence from one species grown in a classhouse pot to another species grown in the field 1\vill perhaps never be accomplished other than by iempiricaJ correlation, but environmental controls !will need to be included in this correlation. ! :2) The variation between replicates (blocks) iu eld trials as an aid to soil science. In glasshouse rials, care is taken to ensure uniform soil and, in articular, a regular moisture regime between aU eatments, so that it is to be expected that the ariation between blocks will be much less than in he field. When analysis of field trials has shown soil (sub-site) differences between blocks of treatments, the principal ambition of experimentalists has been to ensure that this did not invalidate the statistical response differences between treatments. This would now seem the ideal opportunity to describe these sub-site soils in detail and to lay out either a physically controlled or a physically described trial, -to find those soil properties contributing to the observed growth responses between blocks or sub- -sites. Bearing in mind that the original soil was almost certainly chosen for its superficial uniformity, this would lead to the quick elimination, at least, of those soil properties not relevant to crop perfor- mance. {3) The limited utility of soil surveys-except in the special circumstances of coutrolled or relatively uniform plant environment. It is a common experience to :find different soils (in the soil-survey sense) behaving very similarly in the field, and the reverse situation of non-distinguishable soils (again in the survey sense) behaving very differently in terms of plant behaviour, when lying contiguously in the eld. This suggests, that, until the criteria control- ling crop performance are identified (and this will vary for each crop, the biological paran1eter chosen, and the degree of maturity at which the plant is examined) detailed soil surveys can only be of limited use. In regions, where the environment is very similar seasonally or where the environment can be largely controlled, as in irrigation areas, there will be more chance rif identifying poor and good soils, but generally this will be done by studying biological response and its correlation with miscellaneous soil parameters, not vice versa. responses, so the relationship between arbitrary extracts, whether chemical or microbiological, and field trials will be even more tenuous unless the field environment remains sensibly constant. The corre- lations between arbitrary extracts and pot trials where there is no water stress is always found to be higher than in the field where water is uncontrolled. The successful correlation, in Northern Europe and New Zealand, between extracts and field responses is almost certainly connected with the comparatively high degree of reliability in rainfall, and the fact that some crops are virtually grown at field-capacity moisture content for a large part of their life cycle. (5) The misuse of a common test species for assessing comparative soil fertility in several regions. A plant type chosen for its physiological vigour in a certain environment will be able to detect differences between soils in that optimum environment. If the environment is significantly altered from the optimum for expression of the genetic capabilities of the type, the principal response of the plant will be to the change in its environment rather than to soil treatment by fertilizers. Field trials conducted with the same test species, in an attempt to compare the fertility of soils in different regions, will more reflect the difference between the environments than the difference in field response to fertilizer treat- ments. There is no reason why there should be a common measure of fertility; the soils of Northern Europe, used for successful growing of rye, have a fair fertility as measured by rye grain in that environment, but would have a poor fertility as measured by wheat. (6) poor correlatiOn between plant analyses and soil deficiencies. The level of a given nutrient element in a healthy field crop varies widely during the season and, furthermore, the magnitude of the variation differs in different parts of the plant. Some of this variation can be correlated qualita- tively with meteorological data, since plant physio- logists are aware that "bright", "dull", "cool" or "hot" weather influences the production and persistence of visual deficiency symptoms and the chemical constitution of plants. The majority of correlations obtained between total plant analyses (or arbitrary extracts of plants) and field observa- tions of productiveness or visual symptoms of nutrient deficiencies have lacked precision, and attempts to set "critical levels" have given variable limits of qualitative use in diagnosis. CONCLUSION The interpretation of many field trials involving biological parameters would appear, after a century, to have come to a stalemate for Jack of recorded "control" or environmental data. The establish- ment of further field, fertilizer and soil-amelioration (4) The poor correlation between arbitrary laboratory trials, glasshouse trials, assay procedures, and chemical extracts or micro-biological assays and field detailed soil surveys based an non-biological para- responses. Just as glasshouse trials at near-optimum meters can only lead to a limited advance in our moisture regimes are not easily related with field appreciation of soil-plant relationships. Until we 309 use a more comprehensive style of experimentation and recording to determine the principal soil, site and micro-meteorological parameters controlling biological response we cannot elucidate the impor- tance of, and the inter-relationship between, these factors. It is not easy to define the cost of field trials, as the salaries of officers, who have other duties as well as those concerned with trials, are involved, whether directly in the field, or in laboratories and offices. There is no doubt that in alone the total annual charge is in excess of A1 million: and our argument indicates that the majority of tb; information so obtained is either already known or not capable of quantitative application because of lack of data to interpret it. This analysis bas, of necessity, been limited to the. general field of soil science and acknowledgment must be made to the many agricultural scientists. who have helped to crysta1Iize the opinions expressed here. Printing Work.r, York &: London-54165 310
Soil Structure/Soil Biota Interrelationships: International Workshop on Methods of Research on Soil Structure/Soil Biota Interrelationships, Held at the International Agricultural Centre, Wageningen, The Netherlands, 24-28 November 1991