Copyright is owned by the Author of the thesis. Permission is given for a copy to be downloaded by an individual for the purpose of research and private study only. The thesis may not be reproduced elsewhere without the permission of the Author. A QUANTITATIVE ANALYSIS OF THE VARIABILITY IN THE ACTIVITY OF NITRIFYING ORGANISMS IN A SOIL UNDER PASTURE A thesis presented in partial fulfi lment of the requirements for the degree of Doctor of Phi losophy at Massey University Robert G.V. Bramley 1 9 89 i ABSTRACT variability in the inputs , outputs and transformations of mineral N under field conditions makes the predictive modelling of the leaching of soil nitrate very diff icult . In an attempt to understand and quantify this variability , the activity of nitrifying organisms in the Tokomaru silt loam ( a Typic fragiaqualf ) under pasture was measured using a short-term nitrif ication assay ( SNA ) . Spatial dependence of the variabi l i ty in SNA was examined using geostatistical methods , and the ef fect on SNA of soil pH change through l iming , and of seasonal changes in soil temperature and moisture were invest igated . Nitr i f ier activity and associated soi l properties such as the amount of exchangeable ammonium and the soi l ni trate concentration , were found to decrease in value with depth between 0-24 cm . The greatest decrease in SNA was observed between 0-9 cm depth , but due to the need for suf f icient quantities of soi l to make SNA measurements , and the desire to avoid the possibi l ity of inhibitory ef fects of grass roots on nitrif ication , soil was sampled from the 3 -9 cm depth range for the bulk of the work reported here . Results indicated that the technique of sieving and mixing samples was s at is factory f or removing depth-dependence from the results for spatial variability and other analyses . The spatial variability of SNA , soi l No3 - , soi l moisture content and the pH of the SNA incubation , which was assumed to approximate the f ield soil pH , was investigated over areas of 9 m2 and 625 m2 us ing a regular 1 1 x 1 1 square grid sampling design with minimum sample separations of 30 cm and 2 . 5 m respectively . However , the results of these analyses proved inconclusive , apparently due to the lack of samples separated. by lags that were suf f iciently short in relation to the overal l dimensions of the sampl ing area . Accordingly , spatial analysis of the above properties , together with exchangeable ammonium , was carried out over 625 m2 using a n es ted sampling design that permitted an adequate number of observation pain t s a t l ags r anging f rom 1 2 . 5 cm to 2 5 m . Thi s des i gn was a considerable improvement on the regular square design , although i t had a number of shortcomings , notably bias caused in the estimation of the sample vari ance due to the nes ting o f a large number of data points ii within a small area , and bias caused in the estimation of values of the semiva r i ance a t some l ags- due to m i s s ing sampl ing points at some positions in the sampling grid . The va lues o f SNA , N0 3 - and exchangeab le ammonium were all highly variable and conformed to lognormal distributions . The range of spatial dependence in the variabil ity of SNA , soil N03- and incubation pH was 2 . 4 , 5 . 4 and 6 . 1 m respectively . Exchangeable ammonium , SNA , soil No3- and incubation pH varied isotropically within the sampling area but Ex NH4• showed no spatial dependence . Soi l moisture content was strongly anisotropic , and showed no spatial dependence in one direction , but clear evidence o f dr i f t in a perpendicular direc t i on . These results are d i scus sed in r e l a t ion to t he mos t e f f i c ient s ampl ing strategy for estimation of the mean field N03 - concentration . I t was concluded that suf f icient small localized clusters of samples should be taken to give a low standard error of the mean , with each cluster separated by at least 5 m . In the case of the Tokomaru silt loam , 20 clusters , each comprising 5 samples ( bulked ) , would be required for estimation of the mean f ield nitrate concentration with 9 5 % probability of 0 being within ± 5% of �' the true mean . This represents a l arge sampling ef fort . The a c t iv i t y o f n i t r i f iers was s tudied in relation to soil pH and seasonal changes in soil moisture and temperature over two consecutive years in an attempt to explain the spatial variabil i ty in SNA values . The pH optimum for nitrif ier activity ( pHopt ) was def ined for four variates of the Tokomaru s i l t loam with dif ferent l iming hi stories . Values of pHopt which varied between the four soi ls in the range 5 . 92-6 . 4 5 did not vary markedly w i th s eason , and i t was found that the form of the rela t ionship between SNA and pH remained constant wi th t ime . I t was further observed that the addit ion of l ime in 1 9 87 had the effect of rais ing the mean soil pH and pHopt in previously-unlimed soi l , but had negligible ef fect on either the soi l pH or pHopt in soi l that had been l imed in 1 982 . The significance of heterotrophic relative to autotrophic nitrif ication could not be discerned . iii No signi f icant relationships could be found for the four soils between soil pH , pHopt , SNA, soil moisture content and soi l temperature at 3 0 cm depth . Va lues of SNA ( pmol N g-1 so i l h- 1 ) at pHopt ( SNAopt) were calcul ated from equations f i t ted t o plots of SNA vs . the pH of SNA incuba t ion , and these show a more obvious seasonal trend . SNA va lues calculated for the prevail ing soi l pH (SNApH) were never very dif ferent from values of SNAopt and follow a 1 : 1 relationship over a range of values from 0 . 0 1 5 -0 . 1 1 0 pmol g-1 h-1 ; that i s , the nitrifier activity in the s o i l , i rre spect ive o f va ria t ions t hat were ra ndom ( unknown inf luences ) or associa ted wi th seasonal var iables ( temperature and moisture ) , was near the optimum with respect to the soil pH a t the time of sampling . The ef fect of soil moisture variat ion on nitrif ier activity was further invest igated in an experiment in which soil samples were stored for 1 2 4 days a t dif ferent soi l moisture tensions . The optimum moisture conditions for nitrifier activity in the Tokomaru s i l t loam prevailed at pF 3 . 39 . However , this optimum was less clearly def ined than was the pHopt • S ince the soil moisture status changes considerably with season, whi lst soi l pH does not , it was concluded that nitrif iers were more tolerant of changes in pF than changes in pH . Comparison of these with published results indicates that not only is the soi l n i trif ier population dynamic , and changes in response to changes in its environment , but the degree to which nitrif ier activity i s a f fected by various soi l properties is soi l-specific . It is therefore concluded that the spatial variability of nitrif ier activity will also be soi l speci f ic , and that dif ferent soils are likely to have dif ferent ranges o f spatial dependence for the parameters of mineral N. Furthermorer the fact that SNA is not the only factor governing the soi l No3- concentration , and that other f actors such as plant uptake and leaching are also importan t , indicates that SNA variabil ity i s not necessari ly a good e s t i m ator of s o i l N03- var i abi l i ty . This conclus ion i s certainly supported by the geostatistical aspects of this work . iv ACKNOWLEDGEMENTS I would like to thank the fol lowing people for their part in seeing this proj ect through to completion : P r o f es sor R . E . Whi te f or h i s superv i s ion and guidance throughout . The support given in l oco paren ti s by Bob and his w i fe , Annette , on arrival in this s trange ( ! ) country and their friendship thereafter has been very greatly appreciated . Dr A . N . Macgregor for his supervision . Drs D . R . Scotter , P . R . Darrah , A . B . McBratney and A . Swi f t , and Messrs L . D . Currie, M . Eggels , Mrs H . Murphy and Mrs A . Rouse for their assistance at various stages of the proj ect . The s ta f f and postgraduate s tudents of the S o i l S cience Department for providing the relaxed and friendly atmosphere which made l i fe at Massey so enjoyable . The Vice-Chancellor of Massey University , Dr T . N . M . Waters , for making the necessary funds available for this work . Finally I would l ike to thank my parents , John and Rosalind Bramley , for providing all that made starting this work possible , and Jo Tomp ki ns , who se suppor t and af f e ct i o n was the much needed inspiration for its completion. V A QUANTITATIVE ANALYSIS OF THE VARIABILITY IN THE ACTIVITY OF NITRIFYING ORGANISMS IN A SOIL UNDER PASTURE TABLE OF CONTENTS Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv Table of contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v List o f Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvi SECTION I . INTRODUCTION CHAPTER 1. INTRODUCTION AND AIMS OF THE PROJECT . . . . . . . . . . . . . . . . 1 i . Background to the proj ect - the nitrate leaching problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 i i . Nitrate leaching models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 i i i . The aim of the proj ect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 CHAPTER 2 . vi NITRIFICATION IN SOILS: A REVIEW . . . . . . . . . . . . . . . . . . . . 7 i . The factors af fecting nitri f ication and mineral ization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 i i . Modell ing of nitrif ication . . . . . . . . . . . . . . . . . . . . . . . . . 1 7 i i i . A comment on measured nitrif ication rates . . . . . . . . . . 2 3 iv . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 4 CHAPTER 3 . A THEORETICAL CONSIDERATION OF SPATIALLY DEPENDENT VARIABILITY . . . . . . . . . . -. . . . . . . . . . . . . . . . . . . . 2 5 i . Why do we need geostatistics ? . . . . . . . . . . . . . . . . . . . . . 2 5 i i . Some preliminary data analysis using class ical statist ics . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . 28 i i i . Stationarity and the semi-variance . . . . . . . · . . . . . . . . . . 3 2 i v . The variogram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 5 v . Spatial variation in two dimensions . . . . . . . . . . . . . . . . 40 vi . Other variogram models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 7 vii . Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 1 vi i CHAPTER 4. EXPERIMENTAL METHODOLOGY . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2 A . THE SHORT-TERM NITRIFICATION ASSAY . . . . . . . . . . . . . . . . . . . . . . . 5 2 i . Selection of incubation media for SNA analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 3 ii . Linearity of nitrification rate in the Tokomaru silt loam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 4 i i i . Selection o f ammonium substrate concentration for SNA analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 8 iv . Analysis of exchangeable ammonium . . . . . . . . . . . . . . . . . . 6 1 B . FIELD SAMPLING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 i . S ite details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2 i i . Soil sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 3 i i i . Correlation between moisture contents o f sieved and unsieved soil . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 3 C . STORAGE OF SAMPLES PRIOR TO SNA MEASUREMENTS . . . . : . . . . . . . . 6 7 i . Ef fects of drying and storage on mineral nitrogen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 i i . Ef fects of drying and storage on soil biomass . . . . . . 69 D . CONCLUS IONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 6 vii i SECTION I I . AN ANALYSIS OF SPATIAL VARIABILITY IN NITRIFIER 'ACTIVITY CHAPTER 5 . VARIABILITY IN NITRIFIER ACTIVITY WITH DEPTH AND DISTANCE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 8 A . DEPTH DEPENDENT VARIABILITY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 i . Methods and Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 i i . Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 1 i i i . Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 3 B . SPATIALLY DEPENDENT VARIABILITY . . . . . , . . . . . . . . . . , . . . . . . . . . 9 6 i . Methods and Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 7 i i . Results . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 9 8 i i i . Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 08 CHAPTER 6 . SPATIAL VARIABILITY OF NITRIFIER ACTIVITY - A MORE REFINED ANALYSIS . . . . . . , . . . . . . . . . . . . . . . . . . . . 1 1 6 i . Methods and Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2 1 i i . Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2 1 i i i . Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2 5 iv . Conclusions and recommendations for future sampling strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 4 0 SECTION I I I . FACTORS AFFECTING NITRIFIER ACTIVITY ix CHAPTER 7 . THE EFFECT OF pH, MOISTURE AND TEMPERATURE ON NITRIFIER ACTIVITY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 4 4 i . Methods and Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 4 4 i i . Results . . . . . . . . . . . . . . . ; . . . ; . . . . ·. ; . . . . . . . . . . . . . . . . . 1 4 9 i i i . Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 68 iv . Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 7 6 CHAPTER 8 . A FURTHER INVESTIGATION OF THE EFFECT OF MOISTURE ON NITRIFIER ACTIVITY . . . . . . . . . . . . . . . . . . . . 1 7 7 i . Methods and Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 7 8 i i . Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 80 i i i . Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 9 3 CHAPTER 9 . HETEROTROPHIC NITRIFICATION - EXACTLY WHAT HAS BEEN MEASURED BY THE SNA ? . . . . . . . . . . . . . . . . . . . . 1 98 i . Methods and Materials . . . . . . . . . . . . . . . . . . . . . , . . . . . . . 2 0 1 i i . Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 0 3 i i i . Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 0 9 iv . Conclus ion . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . 2 1 2 X SECTION IV . CHAPTER 10. GENERAL DISCUSSION, CONCLUSIONS AND RECOMMENDATIONS FOR FURTHER WORK . . . . . . . . . . , . . . . . . . 2 1 3 A . A CRITIQUE OF THE GEOSTATISTICAL TECHNIQUES USED FOR THE ASSESSMENT OF SPATIAL VARIABILITY IN NITRIFIER ACTIVITY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1 3 i . Problems caused by anisotropy . . . . . . . . . . . . . . . . . . . . . 2 1 4 i i . Problems caused by changes in the sample variance and variation in the variogram model with time and the scale of sampling . . . . . . . . . . . . . . . 2 1 8 i i i . The relationship between the sill and the sample variance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225 iv . Crossvariogram analysi s . . . . . . . . . . . . . . . . . . . : . . . . . . . 2 3 0 B . SOME CONCLUDING COMMENTS ON VARIABILITY IN NITRIFIER ACTIVITY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 3 2 REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 3 5 APPENDIX I . COMPUTER PROGRAMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 52 GAMMAH . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253 COVGM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 58 Figure 3 . 1 Figure 3 . 2 Figure 3 . 3 Figure 3 . 4 Figure 3 . 5 Figure 3 . 6 Figure 3 . 7 Figure 3 . 8 Figure 3 . 9 Figure 3 . 1 0 x i LIST OF FIGURES Distribution of a set of 2 5 pH data . . . . . . . . . . . . . . . . . . . . . 2 9 Change in the mean and variance with increasing sample number for a set of 25 pH data . . . . . . . . . . . . . . . . . . . 3 1 Experimental variogram for the 2 5 pH. data , and the number of pairs of points separated by each lag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 6 A Change in the value of ¥ ( h ) as the number of pairs of data points used to calculate it increases . . . . . . . . . . . 3 8 Experimental variogram for soi l pH with linear models f itted by ordinary and least squares optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 9 1 2 1 equally spaced pH data sampled from a regular 1 1 x 1 1 square grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1 Distribution of the 1 2 1 pH data shown in Figure 3 . 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2 Change i n the mean and variance with increasing sample number for the set of 1 2 1 pH data . . . . . . . . . . . . . . . . 4 4 Change i n the number o f pairs o f points separated by each lag as the length of the grid side increases from 1 to 1 0 lag uni ts . . . . . . . . . . . . . . . . . . . . . . . . 4 5 Development o f an experimental variogram a s the number of samples used to estimate it increases as the length of the grid side increases from 2 ( 4 samples ) to 1 0 lag units ( 1 2 1 samples ) . . . . . . . . . . . . . . . 4 6 Figure 3 . 1 1 Figure 4 . 1 Figure 4 . 2 Figure 4 . 3 Figure 4 . 4 Figure 4 . 5 Figure 5 . 1 Figure 5 . 2 Figure 5 . 3 Experimental variogram for the 1 2 1 pH data f itted with a spherical model by weighted least squares x i i optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 0 Linearity o f nitrification rate over 1 9 hours i n the Tokomaru s i l t loam a t 2 2 oc . . . . . . . . . . . . . . . . . . . . . . . . . 5 6 Mean monthly weather data ( 1 928- 1 980 ) for the D . S . I . R Grasslands weather station , Palmerston North ( N . Z . Meteorological service , 1 98 3 ) . . . . . . . . . . . . . . . 6 4 Correlation between the gravimetric moisture contents of s ieved ( < 2mm ) and unsieved Tokomaru s i l t loam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 pH optima curves for nitri fier activity in fresh soil and soil that had been stored for 3 weeks . . . . . . . . . . 7 4 Common pH optimum curve fi tted to SNA data for fresh and stored soil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 5 Change i n bulk density and mean volumetric moisture content with depth in the Tokomaru silt loam sampled in mid May . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 2 Depth profiles of ( a ) SNA , ( b ) No3 - , ( c ) Ex-NH4 · , ( d ) incubation pH , ( e ) total carbon , ( f ) total nitrogen , ( g ) C/N ratio , ( h ) total phosphorus , and ( i ) % mineral N in the Tokomaru �ilt loam. sampled in mid May . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 4 Distribution with depth of total carbon , ni trogen , phosphorus , C/N ratio and SNA expressed as a % of their maximum values . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . 92 Figure 5 . 4 Figure 5 . 5 Figure 5 . 6 Figure 5 . 7 Figure 6 . 1 Figure 6 . 2 Figure 6 . 3 Distribution of ( a ) SNA , ( b ) No3 - , ( c ) incubation pH and ( d ) gravimetric moisture content sampled on a regular 1 1 x 1 1 square grid between 3-9 cm over xi i i 6 2 5 m2 • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 99 Experimental variograms of ( a ) SNA , ( b ) No3- , ( c ) incubation pH and ( d ) moisture content sampled on a regular 1 1 x 1 1 square grid between 3-9 cm over 625 m2 with models f itted by weighted least squares optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 0 2 Distribution of ( a ) SNA , ( b ) No3 - , ( c ) incubation pH and ( d ) gravimetric moisture content sampled on a regular 1 1 x 1 1 square grid between 3-9 cm over 9 m2 • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 06 Experimental variograms of (a ) SNA , (b ) N03 - , (c ) incubation pH and ( d ) moisture content sampled on a regular 1 1 x 1 1 square grid between 3 -9 cm over 9 m2 with models fi tted by weighted least squares optimization . . . . . . . . . . . . . . . : . . . . . . . . . : . . . . . . . . . . . . . . . . . 1 09 Sampling design for a nested spatial analysis over 625 m2 • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1 2 0 Distribution of ( a ) SNA , ( b ) N03 - , ( c ) Ex-NH4 · ( d ) incubation pH and ( e ) gravimetric moisture content sampled between 3- 1 2 cm over 625 m2 using a nested sampling strategy ( Figure 6 . 1 ) . . . . . . . . . . . . . . . . . . . . . . . . . 1 22 Experimental variograms of ( a ) SNA , ( b ) No3- , ( c ) Ex-NH4. ( d ) incubation pH and ( e ) gravimetric moisture content sampled between 3 - 1 2 cm over 625 m2 using a nested sampling strategy ( Figure 6 . 1 ) with models f itted by weighted least squares optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 26 Figure 6 . 4 Figure 7 . 1 Figure 7 . 2 Figure 7 . 3 Figure 7 . 4 Figure 7 . 5 Figure 7 . 6 Figure 7 . 7 Figure 8 . 1 F igure 8 . 2 Crossvariograms of ( a ) SNA and incubation pH and ( b ) SNA and No3- sampled between 3 - 1 2 cm over 625 m2 using a nested sampling stategy ( Figure 6 . 1 ) with l inear models f itted by weighted least xiv squares optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 3 7 Change in the pH with time of suspensions of soil in agar solution fol lowing addition of acid or alkali . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 4 7 Mean monthly temperatures for the years 1 986/87 and 1 987/88 with sine curves f itted by ordinary least squares optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 48 pH optima curves for nitri fier activity in soi ls T and TL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . ' . . . . . . . 1 5 0 pH optima curves for nitrif ier activity in soi ls TX and TLX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 5 6 Seasonal variation in pH , pHopt1 SNApH' SNAopt, soil moisture content and soil temperature in ( a ) soil T , ( b ) soi l TL , ( c ) soi l TX and ( d ) soi l TLX . . . . . . 1 62 Relationship between SNAopt and SNApH for soils T , TL , TX and TLX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 7 0 Relationship between pHopt and soi l pH for a range of soi ls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 7 1 Moisture characteristic curve for the Tokomaru silt loam ( s ieved < 2 mm ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 8 1 Levels of ( a ) soil moisture stress , ( b ) nitrif ier activity , ( c ) No3- , ( d ) Ex-NH4· , ( e ) gravimetric soi l moisture content and ( f ) incubation pH in the eight samples s tored for 1 2 4 days at di f ferent levels of soil moisture stress . . . . . . . . . . . . . . . . . . . . . . . . . 1 83 Figure 8 . 3 Figure 9 . 1 Figure 9 . 2 Figure 9 . 3 Figure 1 0 . 1 Figure 1 0 . 2 Figure 1 0 . 3 XV pF optimum curve for nitri f ier activity in the Tokomaru silt loam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 92 ( a ) Relative nitrification rate in the Tokomaru s i l t loam at dif ferent pH and ( b ) a pH optima curve for nitrif ier activity in the Patua soil . . . . . . . . . 200 Nitrif ier activities in soi l sampled in June incubated with a range of N-substrates . . . . . . . . . . . . . . . . . 204 Nitri f ier activities in soil sampled in October incubated with a range of N-substrates . . . . . . . . . . . . . . . . . 207 Experimental variogram of incubation pH of soi l sampled between 3 - 1 2 cm over 625 m2 using a nested sampling design ( Figure 6 . 1 ) with separate models f itted by weighted least squares optimization for the north-south and-Bast-west directions . . . . . . . . . . . . . _ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1 5 The main grid of the nested sampling design ( Figure 6 . 1 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227 Distribution of the ln transformed values of initial No3 - sampled betweeri 3 - 1 � cm . depth over 625 m2 us ing a nested sampling design ( Figure 6 . 1 ) . . . . . 229 Table 3 . 1 Table 4 . 1 Table 5 . 1 Table 6 . 1 Table 6 . 2 Table 7 . 1 Table 7 . 2 Table 8 . 1 Table 8 . 2 Table 1 0 . 1 xvi LIST OF TABLES 2 5 values of soi l pH measured at equal spacings along a transect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 Ef fect of ( NH4 ) 2S04 substrate concentration on SNA value . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 Correlation matrix for a range of soi l properties measured at di f ferent sampling depths . . . . . . . . . . . . . . . . . . . 9 1 The number o f pairs o f points separated by a given lag ( h ) under the two sampling designs used for spatial analysis over 625 m2 assuming isotropic data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1 8 Sample variance ( s2 ) and the parameters of models f itted by weighted least squares optimization to the experimental variograms of SNA , initial N03�, Ex-NH4 · , incubat ion pH and soil moisture content . . . . . . . 1 3 1 Summary of SNA results for soils T , TL , TX, and TLX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 67 Ef fect of l iming on nitrif ier activity in the Tokomaru Silt Loam ( 3 -9 cm depth ) under pasture . . . . . . . . 1 7 2 Mean values of SNA ( � mol N g- 1 h- 1 ) , initial No3 - and Ex-NH4• ( � mol N g- 1 ) in soil samples kept for 1 2 4 days at different moisture tensions and the s igni f icance of di fferences between the means . . . . . . . . . . 1 9 0 Summary of results for the suggested optimum moisture tension for nitrification . . . . . . . . . . . . . . . . . . . . . 1 9 6 Summary of results for the three spatial analyses of SNA and incubation pH . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222 SECTION I . INTRODUCTION CHAPTER 1 INTRODUCTION AND THE AIMS OF THE PROJECT i . Background to the project - the nitrate leaching problem 1 The annual world consumption of fert ilizer nitrogen has been variously estimated at 60 M .t ( Hauck, 1 9 88 ) and 73 M t ( Douglas & Cochrane , 1 989 ) . Douglas and Cochrane ( 1 989 ) estimated that an increase in consumption of fertil izer nitrogen of 1 7 - 1 8 % was needed annually to maintain world food supp l y , w h i l s t H auck ( 1 9 8 8 ) more cons erva t i vely e s t imated that consumption by the year 2000 would be 1 00 M t . On top of this , 9 0 M t of N i s added to agricul tural and pastoral systems annua l l y through biological f i xation of nitrogen ( White, 1 989b) which comprises 79 % by volume o f the Earth's atmosphere . The distribution on a global scale of this add ition of fertili zer N is by no means uni form , 58 % of world consumption being in developed western countries , 35 % in countries with central l y con trol led economies , and 7 % in the developing countries . Although the N-fertilizer market is stable in some parts of the world such as Europe , consumption in others is increasing rapidly - in China for example , by as much as 27 % for the year 1 987/88 ( Douglas & Cochrane , 1 989 ) . Against this background of increasing and uneven N consumption , the envi ronmenta l cons equences o f high f er t i l i zer app lication , in particular , are gaining increasing prominence in the publi� eye , to the extent that in Wes tern Europe , the eutroph ication o f s treams and waterways and nitrate pollution of potable water supplies due to leaching has become something of a cause celebre. Figures for N consumption and nitrate leaching in New Zealand are scarce , although the use of fertilizer-N in New Zealand i s only sma l l , being t ied to user-prosperity more than any other factor ( Douglas & Cochrane , 1 989 ) . Indeed , most N . Z bul l beef farmers , for example , do not as a rule apply N fert i l i z er to their pasture , and tend only to apply i t as a rescue 2 mechanism in t imes of low production ( Baars e t a l . , 1 989 ) . In contrast , during 1 9 78 in Britain , the total input of new nitrogen to agriculture was about 2 M t for the year, comprising 0 . 2 - 0 . 4 M t f rom biological N2 f ixation , 1 . 1 5 M t as fertil izer, at least 0 . 2 8 M t f rom precipitation and 0 . 1 8 M t from imported animal feedstuffs ( Thomson , 1 985 ) . All of this except the animal feedstuffs ( i . e . 9 1 % ) can be regarded as a direct application to l and ; the feedstuf fs represent an indirect application ( Ball & Ryden, 1 984 ) . In the U . K, the ferti l izer component applied to arab l e c rops i s a l mo s t a l l app l i ed in spring , whi lst pastures are ferti l ized through the summer ai�ell . It is therefore not surprising that the l evel of ni trate in water supplies of ten exceeds the so-ca l led a ccep tabl e concentration . This i s especially the case when one considers that 3 7 % of the grassland in England and Wales now receives more than 2 0 0 kg N ha- 1 , and 5 % receives more than 4 0 0 kg N ha- 1 ( Ryden et a l . , 1 98 4 ) . The acceptable range of N in drinking water according to the World Health Organi sation is 1 1 . 3 -22 . 6 mg N dm-3 ( W.H . O . , 1 9 7 0 ) whi lst the European Community treat 1 1 . 3 mg N dm- 3 as a permissible maximum but have a gui de l evel of 5 . 7 mg N dm- 3 ( E . E . C . , 1 980 ) . Al l these f igures were greatl y exceeded by the concentration of nitrate in water draining from the experimental plots of Ryden et a l . ( 1 984 ) and were somewhat similar to the mean annual concentration of N03-N i n water draining from the l y s ime ters o f Dowde l l e t a l . ( 1 9 8 4 ) of 1 1 . 8 - 2 6 . 7 m g N dm-3, a concentration which was unaffected by the rate of fertilizer application in the range 0 - 1 2 0 kg N ha- 1 • The latter suggests that the source of the leachable nitrate may not in fact be applied fertilizer. Ryden e t a l . ( 1 984 ) measured nitrate leaching from grazed and cut swards , and found that leaching losses of No3 - from the grazed ( uncut ) sward far exceeded ( by a factor of 5 ) the loss from cut swards despite a common input of fertilizer, and also exceeded the losses from arable land . The enhanced nitrate movement observed below grazed pastures was attributed to the return in urine and dung of as much as 90 % of the N in herbage consumed by the cattle . This casts considerable doubt on the previously held view that arable land was the main source of leached No:- (White , 1 98 4 ) . The problem of enhanced leaching due to the presence of ruminants i s e x a cerba ted by t he spat ia l h e t erogen e i ty o f dung and urine deposi tions . The result of this i s that ho tspots ( White , 1 98 4 ) of very 3 high inorganic N occur , and since the salt concentration in the urine is high , plants suffer from scorch . Plant uptake is low as a result and so does not significant ly reduce the pool of inorganic N ( Ball _& Ryden , 1 98 4 ) . Despite the many advances in agricultural practice , both technological and otherwise , the efficiency of use of fertilizer nitrogen by crops and pastures is poor and the consequent loss due to leaching i s l ikely to be high . White ( 1 989b ) noted that wet and dry depositional input� of N are insignif icant in New Zealand, and that the biological input of N is much more important than the fert i l izer input on a kg N ha-• basis . The results of Ryden et a l . ( 1 984 ) and Dowdell et a l . ( 1 98 4 ) suggested that fertilizers applied to arable land may not be the major source of leached nitrate , but that grasslands , especially those grazed by ruminant animals were the source of much of the leached nitrate . One is therefore led to conclude that since the use of N fertil izer in New Zealand is so low , i f a New Zealand s tudy were to show s ignif icant leaching o f No3 - from grasslands , it would be confirmed that the predominant source of leached No3 - was indeed grazed grassland , and the continuing debate on the source o f the leached nitrate could be ended . Thus , there is a very real need for a model ( or models ) for the accurate prediction of nitrate leaching , so that steps may be taken to minimise i ts ef fects . i i . Nitrate leaching models In the absence of a direct application of N03-N fertilizer , the s ize of - the nitrate pool available for leaching in a def ined volume of soil at a g iven time depends on the balance between the nitrif ication rate and the rate of nitrate removal by plants and micro-organisms . The nitrif ication rate is determined by the activity of the ammonium oxidase enzyme system in nitri fying organisms which in soil is affected primarily by the supply of NH 4 - subs trat e , s o i l tempera ture , moisture and pH ( Barber , 1 98 4 ; Gi lmour , 1 98 4 ) . These factors are discussed i n Chapter 2 . Given a pool of l eachable n i t rate , the l eaching process is governed by patterns of r a infa l l and evaporat ion , so i l t ex ture· and s tructure , i rrigat ion management ( i f any ) and land use ( White , 1 988 ) . 4 A f irst attempt at modelling nitrate leach ing was presented by Burns ( 1 980a ) . The model was based on the concept of the effective _roo t ing depth above which all the inorganic N in the soil was considered equally avai lable , and below which , all N was considered t otally unavailable . N03 - release from the soil organic matter was ignored and all of the ammonium fertilizer added was assumed to have been converted to nitrate , except for the condition when most of the drainage occurred within s i x weeks o f fertil izer application , i n which case only 5 0 % of the NH4 . was considered to have been nitrified . In addition , the leaching of nitrate was assumed to occur only over the 'winter period between fert i l izer application and the beginning of March , by which time all top dressings had been applied, that i s , it was assumed that most of the leaching had taken place before there was s igni f icant crop uptake . It was found that the slow release of mineral N from organic sources , and the differences between the distributions of mineralized and freshly applied-N , af fected the rate at which nitrate from the two sources was lost , and since no account was taken of mineral ization , it was clear that a more complex model was required . A s imi lar conclusion was drawn by Burns ( 1 980b) who tested his model against data from published N response experiments in the Netherlands and U.K. Predictions of the inf luence of winter rainfall on the ferti l izer residues in spring tended to overestimate the spring No3 - concentration observed in Dutch experiments , and underestimate that for the English data . However , the predictions were broadly correct , and the dif ference between the predicted and measured ef fects appeared to be caused by errors in the estimation of the amount and distribution of autumn nitrate in the so i l , and by other losses of nitrate ( e . g . by denitrification ) which may have occurred during the winter . Again, a more complex model was required . Bresler and Laufer ( 1 97 4 ) and Bresler e t al . ( 1 9 82 ) used the convec tion dispersion equa tion in conjuction with a nitrate product ion term to model nitrate movement , whilst White ( 1 985a ; 1 985a , b ; 1 987 ; 1 989a ) used the equations developed by Jury ( 1 9 82 ) and modelled nitrate leaching as a stochastic process using transfer functions . The former approach is based on the m echanisms governing the spread ( or di spersion ) o f a so lute inj ected into a liquid moving through homogenous porous media ( Dyson & 5 White , 1 987 ) . The latter approach acknowledges that the various processes governing solute transport are poorly known , especially in heterogeneous media such as soil, but it can be used to make predictions oe nitrate leaching based on probability , providing that the nitrate concentration prior to leaching i s known , and that estimates o f gains or losses o f mineral N by other processes can be made ( White , 1 987 ) . Dyson and White ( 1 98 7 ) compared these two approaches for the leaching of chloride through a s tructured c lay s o il , and on the bas i s o f compar i sons between experimental results and model predictions were unable to conclude that one model approach was better than the o ther . However , since the mechan i sms of solute transport are poorly understood , the transfer function approach.seems preferable from a philosophical point of view . In spite of the relative success of the t ransfer function .model for nitrate leaching ( White , 1 989a ) , spatial and temporal variability in the inputs , outputs and transformations of mineral N under field conditions make the predictive modelling of soil nitrate diff icult ( White & Bramley , 1 98 6 ; White , 1 988 ; 1 98 9 a ) . White ( 1 985a ) found that a change of ± s tandard deviation in the mean No3- concentration in the soil solution had a relatively large effect on the prediction o f the amount of nitrate leached . Thus , the estimate of this ( initial) nitrate concentrat ion is crucial to the success ful use of the model, but as indicated above is made very diff icult by the spatial variability of nitrate in the f ield . White ( 1 985a ) took soil cores over an area of 4 m2 and found that their nitrate concentrations ranged from 30 to 226 � g N cm- 3 , whilst White et a l . ( 1 98 7 ) conducted an elementary spatial analysis of soil nitrate and found that almost all the var iance was short-range , occurring within 0 . 4 m . i ii. The aim o f the project The aim of the work reported in this thesis was to try to quantitatively analyse spatial ( and other ) variability in soil mineral nitrogen , so that the nitrate concentration in the soil solution could be estimated more accurately over relatively large areas ( e . g . a f ield) . The areal average No3- concentration can then be used as an input parameter in f ield-scale 6 ni trate l eaching model s . In a departure from previous work on this problem ( e . g . White e t al . , 1 987 ) , the emphasis was placed on s tudying the system that generates the nitrate in an attempt to understand why the concentrat ion i s so var i able . In the f o l lowing two chapters , the l iterature on nitri fication and its modell ing i s reviewed and a precis of geostatistical theory i s presented . In further chapters , the variabil i ty of nitrifier activity over area , depth and time is investigated , together with the way in which factors . such as pH and soi l moisture content may affect soil nitrate concentrations . 7 CHAPTER 2 NITRIFICATION IN SOILS : A REVIEW The minerali zati on of nitrogen in soil comprises two separate processes - ammoni fi cati on , which is the release of NH4. from the soi l organic matter , and n i tri fi cati on , whi ch is the ox idation of NH4 • to N03-. A review of the b iochemistry o f these processes i s given by Focht and Verstraete ( 1 9 7 7 ) . �espite the pre-occupation o f the work reported in this thesis with nitri f ication , it was considered pertinent to review previous work on the whole mineralization process , in addition to the nitri f ication l iterature , prior to carrying out further research , for two reasons . Firstly , as Addiscott ( 1 983 ) pointed out , it is not possible to measure nitri f ica tion in the soi l using added ammonium substrates in i solation from either ammonification or immobi l i zati on ( the re-conversion of inorganic to organic N) . Secondly , nitrification depends on. a supply of NH4-N substrate , but because the source of N for the ammonification process , the soil organic matter, comprises many substances some of which are onl y l oosely def ined , ammoni f i cat ion has not been extensively researched in its own right . As a result , much of the earl ier work on soi l n i trogen transf orma tions dea lt instead w ith the minera l i za t i on process as a whole ( e . g . Stanford & Smith , 1 9 72 ) . S ince the rate-limiting step in soi l N mineralization is the conversion of organic N to NH4-N, under condit ions of adequate aeration over a broad range of temperatures and soi l moisture contents , soil derived NH4• is oxidized to No3- rapidly enough to prevent NH4• accumulation ( Stanford & Epstein , 1 9 7 4 ; Wild & Cameron , 1 9 80 ; Schmidt , 1 982 ) . N02-, the intermediary in the convers ion of NH4 • to N03- is not normally detected in excess of 1 I 3 ppm ( e . g . Reichman et a l . , 1 9 66 ) , and so the rate of N03- accumulation generally reflects the rate of N mineralization . Thus , in the fol lowing discuss ion both nitrif ication and mineralization as a whole are considered . Rosswa l l ( 1 9 82 ) sugges ted that there was a probl em regarding the ident i f i ca t i on of the spec i f ic funct ions of the various nitrifying bacteria . Ni trosomonas has always been accredited with the conversion of NH4 • to N02-, with Ni trobacter completing the transformation to N03-. 8 Howeve r , Ni trosol obu s and Ni trospora are a l so common in soi ls and Rosswal l ( 1 982 ) c laimed that Nitrosol obus was the dominant NH4 · oxidizer i n agricul tura l s o i l s . S chmidt ( 1 9 82 ) even suggested that methane oxidizing bacteria �ay be involved, despite acknowledging that the only micro-organisms known to be l inked directly to nitrif ication in natural environments are the gram - ve chemosynthetic autotrophs which comprise the family Ni t robact eriaceae. Since the purpose of this study was to i nvest igate the var i abi l i ty and magnitude of the ne t production of l eachable nitrat e , the t ypes and d i f ferences -between · the various nitrifying organisms was considered to be of secondary importance and it was assumed that autotrophic organisms were predominantly responsible for the production o f nitrate in the particular soi l studied ( see Chapter 9 ) . Throughout , the generic term nitrifiers will be used to denote a l l soil nitr ifying organisms ( both autotrophic and heterotrophic ) ; s imilar ly , ammon ifiers wi l l be used t o denote those organisms responsible for ammonification . Barber ( 1 98 4 ) suggested that 2 % of the total soil nitrogen could be mineralized each year . Dowdel l and Webster ( 1 9 84 ) applied '5N-label led fert i l izer to soi l s in lysimeters . From the amounts of fert i l izer derived '5N that remained at the beginning o f each cropping season , they estimated that 5 - 6 % o f the residual '5N applied turned over each year . Shen e t a l . ( 1 982 ) estimated that the amount of nitrogen contained in the soil microbial biomass represented about 6 % of the total soi l N , and that about 3 0 % of the '5N residual from labelled fertilizer applied to the soil was associated with the microbial biomass . Using these f igures , and their own results for the change in res idual '5N , Dowdell and Webster ( 1 984 ) suggested that about 2 0 % of the biomass turned over each year . They d i s t i nguished between more read i l y avai lable organi c mat ter ( microbial biomass and roots ) and the total soi l organic nitrogen , and calculated that i f the estimate of biomass N turnover v1as correct , and the total soil organic matter turned over at the same rate , it would be equivalent to a release of 1 0 0 - 1 2 0 kg N ha- ' per year . The output of total nitrogen from the fertilizer treated lysimeters exceeded inputs by 7 6 - 9 4 kg N ha- ' per year and in the unferti l ized lysimeters by 1 2 9 kg N ha- ' per year . Thus , i f these differences were due to net mineralization, then there is good agreement between them and the N turnover rates estimated from the 1 5N results . i . The factors affecting nitrification and mineral i zation 9 Macduff and White ( 1 985 ) found that the rate o f nitrif ication was never less than the rate of ammonif ication , whi lst Kowalenko ( 1 9 78 ) found that nitri f ication proceeded very rapidly after the addition o f NH4 -substrate with only background levels of extractable-NH4 remaining after a period o f 4 2 days . In view of these f indings , and those detai led above , it is a reasonable assumption that the factors affecting nitri f ication are those affecting mineral ization . A number o f f actors have been ident i f i ed as rate-determining for ammoni f icat i on and nitri f icat ion . These inc lude the amount , type and avai labi l ity of substrate , the s ize o f the ammoni f ier and nitri fier populations , which in the case of the ammonif iers is very diverse and therefore difficult to investigate , and the envi ronmental condit ions under which these organisms l ive ; i . e . mineral nutrients , temperature , aeration , soil moisture content and pH must a l l be at satisfactory levels ( Harmsen & Kolenbrander , 1 96 5 ; Legg & Meisinger, 1 982 ) . Organi c substrate In a series of f ield and laboratory experiments , Campbell and Biederbeck ( 1 9 8 2 ) ident i f i ed the importance o f_ c rop r es idue s f o r m icrob i a l - pro l i f eration . For an appreciable net minera lization o f soi l organic nitrogen to occur , the C : N ratio of the decomposing substrate must be below 2 0 - 2 5 ( Harmsen & Kolenbrander , 1 96 5 ) or 20 ( Barber , 1 984 ) . When plant residues with a C : N ratio o f greater than 2 0 were added to soil , nitrate and ammonium levels in the soi l decreased as micro-organisms used up carbon from the residues , and nitrogen was immobil ized as a result ( Bartholomew , 1 9 65 ) . If the C : N ratio was l ess than 20 , nitrogen was released at the rate of decomposition as micro-organisms decomposed the residues . In the f ield , this rate is usually between 1 and 3 % per year when calculations are based on the total amount of soil organic matter ( Barber , 1 984 ) . Thus , providing soil organic matter has a C : N ratio of l ess than 2 0 , under otherwise non-l imiting soil conditions , the supply of 1 0 ammonium for n itri f ication wi l l depend on the size o f the ammoni f ier population and the amount of decomposable substrate . The effect of the concentration of avai lable ammonium is discussed in Chapter 4 . pH There have been a number o f studies on the ef fect of pH on nitrif ication ( e . g . Frederick , 1 95 6 ; Aleem & Alexander , 1 9 6 0 ; Morri l l & Dawson , 1 9 6 1 ; Dancer et a l . , 1 97 3 ; Darrah e t a l . , 1 98 6b ) , but commonly these have involved either long-term perfusion experiments or pure culture studies of ni trif iers growing under laboratory condit ions . However , Schmidt ( 1 982 ) noted that nitri f ication "proceeds at soil reactions far below the pH l imits observed for the nitrifying bacteria in pure culture" , and that mos t observa t i on s had indicated an "arb i t ra r y lower 1 i m i t " for nitrif ication of pH 4 , with obvious nitrif ication between pH 4 and 6 , and pH independent nitrif ication in the range pH 6 to 8 . The optimum pH for m i nera l i za t i on o f so i l organic n i t rogen was stated by Harmsen and Kolenbrander ( 1 9 6 5 ) to be on the alkal ine side of neutrality , fol lowing an increase of 1 00-1 5 0 0 kg N ha_ , per year when acid sandy soils with h i gh humus c o n t e n t w e r e l i med . D ancer e t a l . ( 1 9 7 3 ) s tudi ed ammonif ication and nitrification over a range of soil pH values using l imed plots in which the soil pH had been constant for a number of years , and found that soil pH did not af fect rates of ammonif ication appreciably but s ignif icantl y affected nitrification ; Weier and Gil li am ( 1 98 6 ) had s imilar results . Frederick ( 1 9 5 6 ) found that there was a marked decrease in the nitr i f ication rate as the pH dropped below neutra l i ty , whilst G i l mo u r ( 1 9 8 4 ) f ound t h a t t h e r e was a 1 in ear increas e in the nitrif ication rate over the range pH 4 . 9 to 7 . 2 , and by setting relative No3- production to unity at pH 7 . 2 , derived the fol lowing equation to describe the effects of pH on the level of soil N03-N: N03- = ( E X pH ) - F ( 2 . 1 ) 1 1 where E and F are constant s to whi ch the values 0 . 3 3 and 1 . 3 6 were assigned respectively . Bhat et a l . ( 1 9 80 ) used 0 . 4 and 1 . 6 suggest ing that such a relationship is unlikely to be the same for a l l soi ls . Whi lst equation ( 2 . 1 ) will ( correctly ) predict the cessation o f nitrification below pH 4 , i t does not al low f or the effects of an a lka l ine pH on n i t r i f icat i on ( Darrah e t a l . , 1 9 8 6b ) . Evidence from the l iterature suggests that pH has a maj or controll ing effect on soi l nitrification but i t a lso appears that nitri f icat ion rates in d i f ferent so i l s will be affected to dif fering degrees by soil pH . The work of Pang et a l . ( 1 9 7 5 ) conf irmed that the di fference in nitri fying capacity amongst soils was related to the initial nitri fier numbers whose activities were affected by the initial soil pH . It therefore seems likely with respect to this study , that variability in the nitri f ication rate may be in part due to variability in the soi l pH . This is investigated in Chapter 7 . Al l the above relates to the ef fects on nitrif ication rates of pH as measured i n bul k so lut ions . L i tt le account has been taken in the l i terature of the possibil ity that the pH of the immediate environment o f the nitri f iers may be rather dif ferent to that of the bulk solut ion . Evidence suggests ( Fletcher , 1 98 5 ) that soil bacteria tend to be attached to solid surfaces and do not just dri ft about in the soi l solution . Keen and Presser ( 1 98 7 ) found that the degree of attachment of Ni trobacter to an anion-exchange resin increased with pH over the range 5 . 5 - 8 . 0 . No such increase was observed when the attachment was to glass coverslips , but attached cel l s grew approximately 20 % faster than free cel ls . Fletcher ( 1 985 ) attributed this to dif ferent H. ion concentrations in the mucilage of attached organisms compared to th� bulk solution , although, why this should be different to the H. concentration in the mucilage of unattached bacteria in the same medium i s unc lear . Nevertheless , a ssuming that n itri f iers exist in the soi l in clusters of cel ls ( Mol ina , 1 98 5 ; Darrah e t a l . , 1 987b - see below ) , and that the clusters of cells are attached to surfaces in the soi l , the idea that nitri fying organisms can generate a pH env ironment d i f f erent to that of the bulk s o i l s o lut ion is acceptab l e . The e f f ects o f change in the pH of bulk solutions on nitrif iers must therefore be due to altering the pH gradient between the bulk soluti on and the microbial muci lage ; when this gradient is steep , the pH of the mucilage may be altered sufficiently to af fect the activity of nitri f iers covered by it . 1 2 Temperature Very l ittle work has l ooked at the effect of temperature on nitrogen mineralization per se a lthough Harmsen and Kolenbrander ( 1 9 6 5 ) attributed the seasona l f luc tuat ion in m inera lization rate to changes in soil temperature . Kowa lenko and Cameron ( 1 9 7 8 ) f ound an increase in soil mineral N in spring and considered that it was due to higher temperatures enabling mineralization of substrates which had presumably accumulated during winter when low temperatures prevailed . Anderson and Purvis ( 1 9 5 5 ) studied the effects of low temperatures on nitrif ication in incubated soi l s , and found that whi lst nitri f ication began sooner in some soils than others fol lowing the addition o f ammonium , and maximum rates varied, the dif ferences tended to decrease with increasing temperature . In all but one of their soi ls , the accumulation of N03 - at least doubled between 5 . 5° and 8 . 3 oc ( 4 2°-4 7 °F ) , a lthough Stanford et al . ( 1 9 7 3 ) found that Q , 0 for the mineralization process as a whole was approximately equal to 2 . Frederick ( 1 9 5 6 ) found the greatest increase in nitri f ication between 7° and 1 5 oc , but noted that temperatures f luctuating in a 2 4 hour cycle general ly resulted in an increased rate of nitrif ication at temperatures below 1 5 . 5 oc . Harmsen and Kolenbrander ( 1 96 5 ) found that nitrif ication was inhibited at the upper end of the mesophil ic range and that none took place above 4 5 oc , whilst below their stated optimum range o f 2 5°- 30 oc , i t decreased s lowly and practically ceased near freezing point . Schmidt ( 1 982 ) stated that co ld and wet soi l s were e f fectively inactive with respect to n i tri f i ca t i on , whi l s t Mahli and W'Gi l l ( 1 982 ) thought it l ikely that microbes in cool c l imates would adapt to those conditions . Schmidt ( 1 982 ) noted that there were indeed geographical di f ferences in the optimum temperature for nitr i f ication - 20°-25 oc in northwes tern U . S . A ; 3 0°- 4 0 oc in southwestern U . S . A and 60 oc in tropical Australia . Mahendrappa et al . ( 1 9 6 6 ) added ( NH4 ) 2S04 to different soi ls from western U . S . A and incubated them at a range of temperatures between 2 0° and 40 oc at 0 . 3 bar moisture tension . In a l l the soi l s from northern regions , nitrification was faster at 20° and 2 5 oc than at 3 5° and 4 0 oc , whi lst the reverse was true in the case of the southern soi l s whi ch nitrif ied f a s t e s t a t 3 5 o c . Under c ond i t i ons where the t emperature was "unfavourable" , nitrite accumulated . These results suggest that microbial populations in different soils w i l l adapt to their speci f ic environmental conditions . The work of Nakos ( 1 98 4 ) further supports this . 1 3 Gilmour ( 1 9 84 ) described the effect o f temperature on nitri f ication by the equation : Kmax = exp { A/T + B } ( 2 . 2 ) where T i s the tempera ture , Kmax i s the zero-order rate constant at opt imum soil moisture , and A and B are constants . The exp ( B ) term is the frequency factor and A is equal to the ratio of the activation energy of the reaction to the gas constant in the original form of the Arrhenius equation . However , Macduff and White ( 1 9 85 ) found that nitri f ication was limited by the supply of NH4. irrespective of temperature and Rosswal l ( 1 9 8 2 ) agreed that this wa s probably the "main factor" control l ing nitrif ication . With respect to experimental technique , it seems l ikely from the above , that when studying soi ls from a temperate cl imate ( such as New Zealand ) , incubation temperature is not going to be crit ical , so long as it is kept constant and extreme temperatures are avoided . Moisture content , aeration and osmotic stress It has been wel l known for many years that the autotrophic nitri f iers are s trictly aerobic organisms (Meiklej ohn , 1 9 5 3 ) , and it is therefore to be expected that the soil moisture content and degree of aeration would be of fundamental importance to nitrif ication . Indeed , Bresler and Laufer ( 1 97 4 ) found that the rate of NH4· oxidation was directly related to the degree of oxygen availability by gaseous dif fusion , which was in turn inversely proportional to soi l moisture content . Much of the work on the ef fects of moisture content on mineralization was carried out in the context of the ef fects of -drying { and s toring ) soils on their rates of mineralization ; this is discussed in Chapter 4 . With respect to n i t r i f i ca t i on , Bres ler e t a l . ( 1 9 8 2 ) stated that under i sothermal conditions when the soil water content varies considerably during inf i ltration , redistribution and evaporation , the e f fect of soil mois ture content on n i trate production wi l l be of prime importance . Yadvinder - S i ngh and Beauchamp { 1 9 8 8 ) f ound that nitr i fier activity 1 4 increased with increasing soil water potential , and S indhu and Cornf ield ( 1 9 6 7a ) found that the optimum moisture content for ammonif ication and nitr i f ication in the soil s they studied was equivalent to 5 0 % of the maximum water holding capacity . Reichman et al . ( 1 9 6 6 ) . found that the rate of both ammoni f icat ion and nitrif ication of soi l N were almost direct ly proportional to the soil water content at suctions between 0 . 2 and 1 5 bars . At 1 5 bars there was still measurable nitrif ication . Dubey ( 1 9 6 8 ) found that the nitri f ication rate in a sandy loam increased as the soi l moisture tension decreased from 1 5 to 2 bars and then decreased at lower tensions , al though marked nitri fication ( presumably heterotrophic ) occurred even under f looded conditions . In contrast , he found practically no di f ference in the nitri f icat ion rate between 1 5 and 0 . 3 bars in a loamy sand . Justice and Smith ( 1 96 2 ) found that at the optimum incubation temperature f or ni tri f icat ion in the ir ca lcareous soi l , the start of nitr i f i cation f al lowing addit ion of substrate was delayed at tensions greater than 7 bars , al though nitrif ication did occur at the permanent wi l t ing point . A maximum level of nitrif ier activity was noted by Mil ler and Johnson ( 1 9 6 4 ) in the range 0 . 1 - 0 . 2 bars , a l though there was variation in microbial behaviour with respect to the ammonifiers - those producing NH4 + a t z ero t ens i on did not function either when under tension, or with more aeration than was found at zero tension . Those producing NH4 + at higher tensions did not function with less aeration . S tanford and Smith ( 1 9 72 ) stated that the moisture content after vacuum suction at 60 cm Hg was near optimal for mineral izat·ion and they regarded the o xygen concentration under these condi t i ons as similarly near optima l . Sanchez ( 1 9 7 6 ) reported that accumulation of N03- in the upper hori zons of some tropical soi l s could be explained by the existence of nitri f ication at soil moisture tensions of 1 5 -80 bars s ince the crumbs of such soi l s can ho ld water a t these very high tensions due to their mi croa ggrega t e s tructure . Al l this sugges t s that the response of nitrif iers to changing moisture stress is s6i l-specific . Khyder and Cho ( 1 9 83 ) measured the partial pressure of 02 in the soil atmosphere at several depths and found that when the air-porosity was 1 0 . 5 % ( 3 0 % moisture ) the boundary between the aerobic and anaerobic soil layers occurred at approximately 20 cm depth , but when the a ir porosi ty was increased to 1 6 % ( 2 5 % moisture ) , thi s boundary occurred at 1 5 4 0 c m depth . i . e . a sma 1 1 change in t h e a i r poro s i ty led t o a considerable change in the position of the boundary between the aerobic and anaerobic zones . It may wel l be therefore , that N f lushes . ( Birch , 1 9 5 8 ; 1 9 60 ) may occur on a micro-scale as the groundwater table rises and fal l s , allowing for s ignif icant oxidation of soi l organic matter below the surface horizon . Gilmour ( 1 984 ) observed a linear decl ine in nitrif ication rate as the soi l moisture content decreased over the range 0 . 2 -0 . 1 2 g g_ , and he described the moisture relations of nitrif ication by the equation : K/Kmax = ( C X WC ) + D ( 2 . 3 ) where Kmax is the maximum nitrification rate , K i s the nitr i f ication rate at a speci f ic gravimetric moisture content , WC , and C and D are constants which Gi lmour ( 1 984 ) found to be equal to 4 . 8 and 0 . 3 respectively . There is no theoretical basis for C and D , however , and it seems l ikely that a l i near model i s inappropri ate to decribe the relationship between moisture content and nitri f ication rate s ince the evidence is that both very high and very low moisture contents are inhibitory to nitri f ication . Thi s is investigated in Chapter 8 . The factors discussed above relate primarily to matric potential effects on nitrif ier act ivity . In addition , the osmotic ef fects of particular s o i l so lutes may be important . The e f fects o f osmo t i c s tress on n i t r i f i ca t ion and m inera l i za t ion have not been extensively studied a lthough Darrah et a l . ( 1 985a ; 1 9 8 6c ; 1 9 87a ) looked at the ef fects of high concentrations of ( NH 4 ) 2S04 and NH4Cl on nitrification rates . This work is discussed in Chapter 4 . Sindhu and Cornfield ( 1 9 67b ) studied the effects of chlorides and sulphates of Na , K , ea and Mg added in solution at concentrations of 0 . 1 -2 . 0 % ( Na equivalent ) on N mineral i zation and nitrification . cl - in concentrations between 0 . 5 and 1 % caused almost complete suppress i on of nitr i f icat ion , but mineralization was only reduced when a concentration of more than % salt was added . S04 2- only reduced mineralization and nitrif ication when added as 2 % Na2S04 . In some cases , both sulphates and chlor ides of all cations �xcept Na resulted in small but s ignificant increases in N mineralization . Whether 1 6 o r not the sodium response i s a toxic effec t i s unclear , but Harris ( 1 98 0 ) noted that Na as NaCl ( along with sucrose ) was the most important solute with respect to osmotic stress in soils . Here it is enough to say that high osmotic stress is inhibitory to nitri fiers and presumably ammoni f i ers too . One might therefore conclude that equa tion ( 2 . 3 ) is either incomplete , or that C and D are in some way dependent on either osmotic stress , degree of aeration or both . Other factors In addition to the factors discussed above which may be thought of as most importan t , there may be other less obvious factors which will i n f l uence n i tr i f icat ion rates in certain s i tua t i ons . For example , Purchase ( 1 9 7 4 ) found that P defi ciency affected nitrif ication t o the extent that N02- would accumulate i f P levels were low enough . Loveless and P a inter ( 1 9 6 7 ) demonstrated that the effect of def iciencies of copper , s odium , ca lcium and magnes ium on the growth of Ni trosomona s europaea were such that the effect of pH was dependent on the metal ion concentrat ion . pH was also found to strongly influence copper toxicity . The l iterature on the ef fects of pH on nitri f ication rates suggests that the nitrif ication rate is highest at neutral or mi ldly alkaline pH ( see above ) . Some of the mi ldly acidic New Zealand soils ( Yel low brown learns ) studied by Steele e t a l . ( 1 980 ) were found t o have surprisingly h igh nitri f ier activity ( 2 -3 � g N g_ , h- ' ) , this activity being of the order o f , or h igher than , that o f so ils of near neutrq, l pH . Sarathchandra ( 1 9 7 8 ) noted that the dominant clay mineral in these soils is al lophane , and thought it probable that at pH 5 . 5 while negative charges on the a l lophane surface retain some NH4+ , the positive charges present may in fact repel H+ ions , establishing a micro-site containing fewer H+ ions than the bulk soi l , and thus a higher pH than the bulk solution . This explanation is based on the assumption that pH 5 . 5 is close to the point of zero charge , that is , at this pH al lophane has approximately equal numbers of positive and negative surface charges ( Sarathchandra , 1 9 78 ) . However , at the point of zero charge the pH measured in water is similar to that measured in salt solution , and so the difference between the pH 17 at adsorpt ion sur faces and the bulk solution is likely to be small . Sarathchandras ' explanation for high ni tri f ier act ivity in all ophanic soi ls may therefore not be correct . ii . Modelling o f Nitri fication At the s implest level , Gilmour ( 1 98 4 ) took equations ( 2 . 1 ) to ( 2 . 3 ) and by combining them and adding an expression for the ef fect of substrate concentration, Nt , ( see Chapter 4 ) , calculated the absolute nitrif ication rate NR , according to the equation : NR { [ exp (A/TB ) ] X [ ( C X WC ) + D ] X [ ( E ll' pH ) - F ] } Nt 0 . 9 5 ( 2 . 4 ) where the symbols are as before . However, this seems little more than an elaborate curve f i tting exercise; in any case , the value of equations ( 2 . 1 ) t o ( 2 . 3 ) is doubtful . C learly something more sophisticated is required . In a l l the models describing the nitrif ication process which are outlined below , i t has been assumed that the oxidations involved are zero-order react ions ( e . g . W"Laren, 1 9 7 6 ) . Addi scot t ( 1 9 8 3 ) quest ioned whether nitri f ication was truly zero-order on the grounds that nitrif ication rates are not independent of the initial NH4 . concentration ( see Chapter 4 ) . Molina et a l . ( 1 9 79 ) suggested that the kinetics of NH4 . oxidation are the resultant averages of pulses of activity from small isolated and asynchronous clusters of NH4 . oxidizers , and a theoretical consideration o f cell clustering ( Darrah e t a l . , 1 9 87b ) supported this . The oxidation o f ammonium around each c luster , which may comprise a few hundred cells ( Mol ina e t a l . ( 1 979 ) , i s very rapid and the exponential ( f irst-order ) mode l applies . Thus , i f the soi l can be regarded as a single large aggregate with only one cluster , in the non-steady state the kinetics of n i t r i f icat ion wi l l be f i r s t - order and wi l l f ollow the kinetics of m i crob i a l growth , and it may be as sumed that the rate of ammonium o x ida t i on w i l l not be cons tant unt i l steric saturation is achieved ( Mo l ina et a l . , 1 979 ) . At this point , steady s tate conditions apply and 1 8 N0 3 - product ion i s constant . Thus , zero- order kine t i cs apply . In addition , Molina et al . ( 1 9 79 ) found that there was no synchronization of the beginning of nitri f ication amongst the aggregates tested , despite the f act that they a l l c ame from the same f i eld s ample ; indeed , some particles took eight weeks to exhibit their nitri fying potential . Thus , even under non-steady state conditions , zero-order kinetics will appear to apply due to the averaging ef fect of cluster asynchronicity . M0Laren ( 1 9 69 ) ignored this problem on the assumption that the whole microbial population will grow unifor�ly until a maximum population is achieved , and that this population continues to carry out nitrif ication with very little multiplication . As outl ined below , this assumption is not really acceptable , a lthough i t enabl ed init ial progress to be made in the modelling of nitri f ication . Burns ( 1 980a ) modelled with a deliberate lack of analysis of N interactions so that the " main e f fects of the s low release of nitrate and the distribution o f rainfall on the leaching of nitr i f ied-N" could be considered . Whether or not this is a helpful approach is debatable . M0Laren ( 1 969 ) modelled the concentrations of NH4 · , N02- and No3 - under cond i t i ons o f m i crob i a l s teady s tate ( i . e . w i t h no growth in the microbial population ) as NH4 · was applied to the top of a soil column by the equations : ( N02- 1 = { k , . ( NH4• ) o } { exp ( -k , x ) - exp ( -k2x ) } ( k2 - k , ) ( 2 . 5 ) ( 2 . 6 ) ( 2 . 7 ) where x is the distance of f low down the colu�n , k , is th� rate of the NH4 · to N02- oxidation divided by the flow rate down the column , k2 is the rate of the N02- to No3 - oxidation divided by the f low rate down the column , and [ NH4 · ] o is the concentration of ammonium in solution applied at the top of the column . M0Laren ( 1 9 7 0 , 1 9 7 1 ) used these equations as inputs to the main model which gave the rate of either oxidation at small substrate concentrations by : -o [ S ] = A¥m + am + ( � � m [ S ] ) ot ( km + [ S l l 1 9 ( 2 . 8 ) where S i s the substrate concentration , m is the biomass , A is a proportional ity constant ( the reciprocal of growth yie ld ) equal to N oxidized per unit weight of biomass synthesized , a is the N oxidized per unit weight of biomass per uni t time for cel l main tenance, � is the amount of enzyme per unit biomass involved in waste me tabol i sm , ¥m = om/ot for growth me tabol ism, � is a proportionality constant and km i s a saturation constant . Thus , the f irst term relates the disappearance of subs trate due to microbial gro wth , the second term provides f or maintenance of the population in the absence of growth , and the third term represents the rate of change of substrate in addition to growth and maintenance , the waste metabolism . According to M0Laren ( 1 9 7 0 ; 1 97 1 ) it is this which provides N03- for plant uptake and may be regarded as occurr ing s i mply because the enzyme system i s present and active . However , this parti tioning of NH4 � oxidation into growth , maintenance and waste metaboli sm i s misleading because the only reason that nitrifying organisms oxidize NH4 • is to gain energy for growth . According to Wild ( 1 988 ) , the Ni trosomonas group of bacteria oxidize 3 5-70 moles of NH4• for every mole of carbon assi milated, and Ni trobacter ox idize 7 0- 1 0 0 moles of N02 - for every mole of carbon assimilated . If we assume that the C : N ratio of nitrifying bacteria is approximately 6 ( Brady , 1 984 ) , it fol lows that the amount of N that is assimilated by the organisms is so small in relation to the amount oxidized , that virtually a l l the N03- produced is available for release to the medium ( i . e . assuming for the sake of argument that the mass of a mole of C and N is approximately the same , 6 moles of C assimi lated gives 420 moles of oxidized N of which only 1 i s retained for protein synthesis ) . Nevertheless , M0Laren ( 1 9 7 0 ) persisted with the a , � and ¥ terms and presented sub-models for a range o f scenarios w i th vary ing degrees of substrate concentra t ion and microbial enrichment . 2 0 M0Laren ( 1 9 7 6 ) acknowledged that , in fact , the steady state situation is never achieved due to ion exchange and leaching of nutrients . In a f lowing solution , hydrodynamic dispersion, equations for which were given by K irkham and Powers ( 1 9 7 2 ) , may a lso be important . Thus , for a microbial oxidation in a column of soil with the substrate moving at a f low rate f, the change in concentration with time is described by the equation ( M0Laren , 1 97 6 ) : o [ S l ot -f (o [ S ] ) + D ( o2 [ S ] - � [ S ] ox ( 2 . 9 ) where D is hydrodynamic dispersion and � [ S ] is some function o f [ S ] that represents change , i . e . loss of substrate by microbial oxidation . M0Laren ( 1 97 6 ) presented equations to describe � [ S ] under conditions of different growth , f low and substrate concentration in terms of a , � and ¥ . Despite the fact that M0Laren ( 1 9 70 ; 1 9 7 1 ; 1 9 76 ) presented a nice theoretical model with respect to these terms , _ they cannot in practice be in any way separated or identi f ied , and as indicated above , there seems to be little j ustif ication in distinguishing between them . An alternat ive treatment was therefore required . The rate of growth of nitrifying biomass , m, was modelled by Darrah e t a l . ( 1 9 85b ) as : om = J,J max { [ C , l t } m ( 2 . 1 0 ) -a t [ C , J t + k .. where J,J max i& the maximum specific growth rate ( h- 1 ) , ka i s an affinity constant ( J.J moles cm-3 ) , and c , is the ammonium concentration in solution, where at any time t : [ c , l t = [ C , l o - I oC2 + J· ( o [ CN l ) ot o t o t ( 2 . 1 1 ) where [ C , ] o is the NH4 . concentration at time zero and C2 is the No3- concentration per unit soil volume . The last term describes the rate of ammonif ication for native soil organic N . 2 1 The rate of No3 - formation is given by the equation ( Da rrah e t al . , 1 98Sb ) : { 1 } {am} Y a t ( 2 . 1 2 ) where Y is the yield constant defined as � g biomass formed per � mole o f NH4 - transformed . Substitution from equation ( 2 . 1 0 ) gives : oC2 = � max { [ C , ] t {m } ( 2 . 1 3 ) ot [ C , ] t + ka Y Then dividing both sides of equation ( 2 . 1 0 ) by Y gives ( Darrah e t al . , 1 9 8Sb ) : o ( m/Y ) = IJ ma.x { [ C , ] t ------ ot [ C , ] t + ka {m } y ( 2 . 1 4 ) and thus the formation of N03 - can be expressed in terms of equations ( 2 . 1 3 ) and ( 2 . 1 4 ) containing the three parameters ( m/Y ) , � ma.>< and k .. . When C , is less than , or of simi lar magnitude to k .. , equations ( 2 . 1 3 ) and ( 2 . 1 4 ) are solved numerically , but where c , is greater than k .. , they can be integrated to give the nitrate formed in terms of the two constants ( mo/Y ) and � max ( Darrah et al . , 1 98Sb ) . U s ing this theory as a bas i s , in addi tion to equations describing dif fusion of NH4- and No3- in soi l ( Darrah et al . , 1 983 ) , adjustments were made to enable the model ling of simultaneous nitri f ication and dif fus ion in soil with respect to the addit ion of ammonium sulphate ( Darrah et al . , 1 98 6a ) , pH ( Darrah et al . , 1 986b ) , and osmotic potential ( Darrah et al . , 1 987a ) . A simplif ication of the model is presented by Darrah et al . ( 1 98 6d ) . 2 2 Models such as those o f M':Laren ( 1 970 ; 1 97 1 ; 1 9 7 6 ) and Darrah e t a l . ( 1 9 8 5b ) require some estimat ion of the size o f the microbial population . Many workers have used the most probable number ( MPN ) technique of estimating microbial populations ( Cochran , 1 9 5 0 ; Schmidt , 1 982 ) . However , this technique shows a high degree of variability and cannot be regarded as accurate ( Schmidt , 1 982 ) . Morri l l and Dawson ( 1 9 6 1 ) initiated the development of an alternative indirect method of measuring the nitri f ier population when they noted that advantage could be " taken of the facts that chemoautotrophs are the maj or , if not the sole agents concerned with N03 - production in nature , and that the oxidation of ammonium and nitrite compounds i s growth linked . Hence by measuring the quantity of ammonium or nitrite-N oxidized , the rate of growth of the respective bacteria can be ascertained . " However , i f the time period over which NH4· oxidation occurs i s l imited , then the assumption can be made that microbial growth is minimal and the nitri fication rate measured is an index of the size and activity of the nitrifier population . By optimizing conditions in terms of substrate , 02 , temperature and moisture , so that each organism can f unc t i on o pt i ma l l y , the n i t r i f ica t ion ra t e in a short - t erm n i tri fi ca t i on assay, SNA , should be an index of the number of organisms presen t . This was the bas is o f the work o f Sarathchandra ( 1 978 ) and Steele e t a l . ( 1 980 ) who found that the main benefit of the SNA was that the short t ime of perfusion or incubation meant that the results were unaf fect ed by microbial pro l i feration , and thus a good ref lection of nitrif ication activity was obtained . The technique employed ip the SNA { Sarathchandra , 1 9 78 ; Steele et a l . , 1 98 0 , Darrah et a l . , 1 986b; 1 987a ) g ives a h i gh degree o f reproduc ib i l i t y ( P . R . Da rrah persona l communication ; see also chapters 4 -9 ) and may thus be used to give useful input data to nitrif ication models . Mol ina { 1 98 5 ) looked at nitri f ication in a completely di f ferent and novel way . As already mentioned , he assumed that nitrification proceeded from pulses of ammonium oxidation generated by microbial clusters , and in this connect i on carr ied out his e xper iments with individual soil micro aggregates rather than soil columns . He noted that for every NH4· ion oxidized , two H. ions are released , resulting in a pH decrease : 2 3 ( 2 . 1 5 ) and therefore used an experimental procedure involving bromothymol blue indicator which al lowed the pH change to be monitored by transmittance through a spectrophotometer . This method may be potentially use ful in invest igat ions of variabil ity of ni tri f ica t i on on a micro - s ca l e , especially in relation to the differences in nitri f ier activity between the inside and outside o f clods . iii . A comment on measured nitrification rates Brandt e t al . ( 1 9 6 3 ) f ound that there were large discrepancies in their results between NH4 . disappearance and No3- accumulation and as a result , had to distinguish between ni tri fi ca ti on as the biological oxidation of reduced forms of N and net ni tri fi ca tion as the observed accumulation o f No3- · With respect t o experimental techniques such a s the SNA , this is of great importance . I t was stated at the beginning of this chapter that it is not possible to measure nitrif ication in soi l in isolation from either mineralization or immobili zat ion ( Addiscott , 1 9 83 ) and it is implied above that the SNA measures net nitrate production irrespective of the mode of production . It is not within the scope of either this thesis or this review to cons id er the vast l i terature on deni tr i f i ca t i o n , immobilization , and volatilization of soil N . However , it is wel l known ( e . g . Starr et a l . , 1 9 7 4 ; Kowalenko , 1 978 ) that nitri f ication can occur simultaneous ly w ith these processes . Indeed , Colbourn et a l . ( 1 9 8 4 ) demonstrated that in a drying soil , nitri f ication was rate-determining for denitri f icat ion . Furthermore , when NH4 • substra tes a re added in incubation exper iments , some may be f i xed by - clays ( Mogi levkina & Lebedeva , 1 982 ) depending on the amount of available NH4 . and the clay percentage ; Darrah e t al . ( 1 985b ) attributed their incomplete N recovery to the fixation of NH 4• by mica-type clays . Thus , until such time that measurements of these various processes can be made in isolation f rom one another , net measurements will have to suffice . There is no indication in the l iterature to suggest that any interpretation is lost as a result . Moreover , if abundant NH4· is added , the f ixation of NH4 · by clays will 2 4 be unimportant ; i f the system is wel l aerated , denitrif ication will be insign i f icant ; if incuba tion times are short so tha t growth o f the o rga n i sms i s m i n ima l , then N03 - i mmob i l i z a t i on s ho u l d a ls o be insignif icant , and the concentrat ion of NH4 • should ensure that the requirements of both the autotrophic nitrif iers and the heterotrophs are satisf ied . iv . Conclusion Overal l , it appears that the variability in nitri fi er activity may be a function of variabilty in the factors discussed in Section i ( above ) , most importantly soi l pH , mo is ture content and availabil ity of NH4 • substrate . Since the ef fects o f the f irst two of these factors appear to be soi l -specific , they clearly merit attention in this study . For a study o f variabil ity in nitrifier activity , the SNA appears to have the most potent ial since it is much quicker than the other techniques , gives readil y reproducible results , and in view of the small amount of soil required ( Darrah et a l . , 1 986b) may be very suitable for spatial studies involving large numbers of samples . In addition , the problem of dif ferent nitrifying species can be ignored with thi s technique ( P . R . Darrah - personal communication ) . 2 5 CHAPTER 3 A THEORETICAL CONSIDERATION OF SPATIALLY DEPENDENT VARIABILITY The main obj ective of the work reported in this thesis was to investigate the v ar iab i l i ty o f ni tr i f i e r act ivi t y . As was i nd icated in the introductory chapter , in addition to investigating this variability in relation to factors which might be expected to control nitrif ication , it was of maj or interest to investigate and quantify the spatial variability of nitrif ier activity so as to improve estimates o f the initial nitrate concentration as an input parameter for nitrate leaching models . This can not be readily done using classical statistics . The following discussion expl ains why this is so , and outlines the means by which spatial variabi lity may be investigated . i . Why do we need geostatistics ? The traditional means of statistical analysis that soil scientists have used to corroborate the hypotheses which inspired their experiments - described by Nielsen ( 1 987 ) as " aggie statistics" - involved calculation of the mean and variance of sets of data col lected from regions , or under conditions , that were perceived to be horr.ogeneous ,. Often this was done , and sti l l is today , with total disregard for the distribution of the data about the mean , and with results explained in terms of cause and ef fect ; when an ef fect could not be attributed to a cause , i t was explained away by an error term ( usually the residual mean square or variance ) which was commonly ascribed to the inadequacy of sampling , the inaccuracy of an analytical technique or simply to random variation . In fact , a large part of such error is most likely to be due to measurements being made in non homogenous areas , but this possibility has either been ignored , or the worker has been ignorant of the possibility of some spatial dependence in the data . Given this background of an experimental ahd statistical sta tus quo, D . R . Nielsen speaking to the Dutch Soil Science Society on their 5Qth anniversary , argued that " . . . . [ incorporated into ] the next page of soil science should be regional ized variable theory . . . . If we are to achieve greater succes s , we mus t take advantage o f spa t i a l and temporal vari ab i l i t y instead of avoiding it . I f we acknowl edge i t s existence , it w i l l enhance our research e fforts even when we subj ect experimental sites to selected treatments . " He commented further : " . . . . The mean value of a soi l property , which we have become so accustomed to seek and appreciat e , may not , in the f inal analysis , be as important as its spatial and temporal variance or the identi f ication and possible signi f icance o f its per turbed values . " ( Nielsen , 1 987 ) . 2 6 Thus , the need has arisen for quantitative spatial and temporal analysis of soil properties , to complement the investigation of cause and ef fect relationships between them . In t he f o l lowing sec t i ons , the rat iona le behi nd geos ta t i s t ics is presented . Since much of the theory is not new and has been extensively detailed in a soil science context previously ( Webster , 1 98 5 ; Trangmar et al . , 1 98 5 ; Ol iver, 1 987 ) , the topic will be developed here in relation to the work presented i n this thes i s , and thus some aspects , such as interpolation by kriging , are omitted . In this discussion , geostatistical techniques are looked at from a dif ferent angle to that presented in the l iterature with the intention of explaining to the uninitiated , and the non-mathema t i c i an in part icular , the poten t i a l o f these power ful statistical tools . For the purpose of this discussion, a data set comprising 2 5 soil pH values wi l l be considered ; the data are l isted in Table 3 . 1 . I t will be assumed that they represent measurements taken along a transect within an area which is assumed to be homogeneous , wi th equa l spacing between samples ; samples No . 1 and 2 5 represent the two ends of the transect . Table 3 . 1 2 5 values of soi l pH measured at equal spacings along a transect Sample pH Sample pH No . No . 5 . 38 1 4 4 . 9 3 2 4 . 38 1 5 4 . 8 0 3 5 . 00 1 6 4 . 88 4 5 . 1 0 1 7 4 . 88 5 4 . 65 1 8 4 . 9 0 6 5 . 2 0 1 9 4 . 9 3 7 4 . 78 2 0 5 . 0 5 8 5 . 38 2 1 4 . 8 5 9 4 . 95 2 2 4 . 8 5 1 0 4 . 78 2 3 4 . 9 5 1 1 4 . 70 2 4 4 . 80 1 2 4 . 5 3 2 5 4 . 8 5 1 3 4 . 68 i i . Some pre l iminary data analysis using classical s tatist ics 2 8 Any kind of analysis requires a starting point , and one useful way to begin to analyse a data set is to f ind out about its distribution ; i . e . one needs to know the mean and variance , and the way the data are distributed about the mean . It i s assumed that the sample has been drawn from a population with true mean � ' and variance o2 ( Clarke , 1 980 ) . The true mean value of a property z , is estimated by P , __ the ari thmetic mean , where ( Clarke , 1 980 ) : A � "' r. Z ( x:�. ) ( 3 . 1 ) n and the population variance o2 is estimated by s 2 , the sample variance , by ( Clarke , 1 980 ) : ( 3 . 2 ) n - 1 where n i s the number of realizations or values of the property Z , and x def ines the location , in cartesian coordinates with i "' 1 , 2 , 3 . . . n , at which the individual value s of the property Z a re observed or measured . Applying these equations to the data in Table 3 . 1 , values of 4 . 89 and 0 . 0 5 1 7 are obt a ined f or the ari thmetic mean and s ampl e variance respectively . White e t a l . ( 1 987 ) noted the dangers of assuming that the s imp le ar ithmetic mean and variance were the best es t ima tes o f the population mean and variance when s ample data are skewed and do not conform to a normal distribution . When this i s the case , an estimate of the ar i thmet i c mean derived f rom the parameters of a l og -norma l distribution, o r one based on Sichel ' s estimator ( Siche l , 1 9 5 2 ) may give better estimates of the population mean . Thus , the distribution of the data must be checked . The 2 5 pH data were grouped into classes ( 0 . 2 pH uni ts wide ) , the f r equency o f each c l a s s normal i z ed , and the resultant distribution plotted ( Figure 3 . 1 ) . This was shown by means of a least-squares f itting Normal ised frequency 2.5 2 1 .5 1 0 .5 0+-��--�----+---�--�----+-��-A�--� 4 4.2 4.4 4.6 4.8 5 pH Figure 3 . 1 Distribution of a set of 2 5 pH data 5.2 5 .4 5 .6 5.8 3 0 procedure ( CFIT - Dept . Soi l Science , Massey University ) t o conform to the normal distribution f( x ) ( R2 = 0 . 9 4 , p< 0 . 1 % ) where ( Clarke , 1 980 ) : f( x ) = ( 3 . 3 ) Here , Ql and "C' 2 are best est imat es of the population mean and s ample variance , IJ and and w i s the mid-value for each pH c lass . 0 is calculated as an average of a l l the values of Z ( xi ) , and it should give a good est imate of the property z , in this case soil pH , when that is measured at any point xi along the transect . We therefore say that the expec t ed val u e of z at any point x i along the transect i s given by ( Webster , 1 985 ) : E [ Z ( xi ) ] = 1J ( 3 . 4 ) where E denotes expectation . Taking this idea to its logical extension and considering just the first point on the transect , it can be argued that i f the value of z at t his f irst point is unknown , i t could be expected to be equal to � which i s an estimate of v in equation ( 3 . 4 ) ; i . e 4 . 89 . Conversely , i f the mean was unknown but the value of z at this point was known , and assuming that the distribution of pH values along ,\ this transect was normal , IJ would be expected to be equal to the value of z i . e . 5 . 38 . Using this argument , and equation ( 3 . 1 ) to calculate the value o f 0 when n i s greater than 1 , the change in the est imated mean and variance with increasing sample number was investigated by starting at one end of the transect and moving along i t , one sample separation at a t ime to include the other points , unti l the whole data set was included . The results are shown in Figure 3 . 2 as a plot of the mean and variance vs . the number o f va lues used to ca lculate them . Since the sample variance of a single real ization is zero , the change in s2 was calculated for n=2 to n=2 5 . By def inition , the variance o2 i s a measure of the scatter or dispersion A of the values of Z ( xi ) about the mean IJ ( Clarke , 1 980 ) . 52 and IJ are also assumed t o be good e st imates of the true variance and mean of the population from which the sample data are drawn . The s ize of s2 tel ls us 5.6 5 .2 Mean 4.8 (�) 4.4 4 0.5 0 .4 Variance 0.3 0 . 1 0�-------+--------�------�--------�------� 0 Figure 3 . 2 5 1 0 1 5 20 25 No. of samples . A 2 Change �n the mean ( � ) and variance ( s ) with i ncreas ing sample number for a set of 2 5 pH data 3 2 1\ something about the precision with which � is measured and we can assume " that the more precisely � is measured , the nearer i t will approach to � · I t is therefore interesting to note from Figure 3 . 2 that as n increases , " the change in the value of lJ calculated for n and n+ 1 realizations of Z ( x.._ ) decreases , to the extent that for values of n greater than 1 5 , there i s very l ittle f luctuation in the estimated value of � relative to the f luctuation when n is less than 9 . That is , the greater the value of n, the lower the value of 52 ( Figure 3 . 2 ) , and consequently the more 1\ precise the est imate of lJ by lJ . This idea of precision will be seen to be important in the fol lowing sections when the spatial distribution of the Z ( x.._ ) i s · considered . In addit ion to the concepts of the mean and vari anc e , i t is a lso an important pre-requisite of spatial analysis to understand the concept of covariance . Supposing that in addition to soi l pH , the C . E . C had been measured at each s i te , x .._ , a long the transect . As part of the data analysi s , it may be useful to have a measure of the correlation between the two properties , denoted here by Z and Y . This can be estimated by the covariance , COV , where ( Clarke , 1 980 ) : ( 3 . 5 ) n - 1 1\ ;\ where l.l z and lJ v are the mean values of the sample data of Z and Y . It is shown below that the concept of covariance is important to geostatistical theory . iii . Stationarity and the semi-variance Although equation ( 3 . 4 ) states that the expected value of Z at any point x .._ is lJ , it is clear from Table 3 . 1 and implicit in Figure 3 . 2 that the value o f z wi ll in fa ct vary from place to place . Thi s would more obviousl y be the case i f , for example , the transect crossed the boundary between two distinct soi l types . Thus , Z ( x.._ ) is cal led a random vari abl e - geostatistics are concerned with identifying its spatial structure . By spatial structure is meant the spatial correlation of the variable with i ts e l f , . which can be described by propert ies of i t s probabi l i ty distribution . The f irst property is the mean , def ined by equation ( 3 . 4 ) ; the second i s the spatial covariance , defined below . 3 3 When the mean value o f z ( x:1. ) does not vary along the transect , the condition of firs t-order sta tionari ty is said to hold ( Webster , 1 985 ) : E [ Z ( x:t. ) l = � = constant ( 3 . 6 ) and i t would be expected that a plot l ike Figure 3 . 2 would show a straight horizontal l ine corresponding to a pH of 4 . 89 . I f equation ( 3 . 6 ) holds , the expected di f ference between any two values of Z ( x:t. ) separated by a distance or l ag, h , would be zero ( Trangmar et al . , 1 985 ) : E [ Z ( x:t. ) - Z ( x;�. + h ) ] = 0 ( 3 . 7 ) If the mean does vary , dri ft is said to be present and the changing value of the mean can be described by the dri ft function , d (x;�. ) ( Starks & Fang , 1 982 ) , and equation ( 3 . 6 ) can be re-written more generally as : ( 3 . 8 ) where w ( x:t. ) is a random function of zero mean and finite , f ixed variance . w ( x1 ) depends on the variation between values of Z ( x:t. ) and Z ( x:t. + h ) , for all values of h . One of the aims o f geostatistics i s to quanti fy the degree o f spatial correlation between the values of Z ( x:t. ) and Z ( x;�. + h ) . This can be done using the concept of covariance , expressed mathematical ly in equation ( 3 . 5 ) . Thus , the spa t i a l covari ance of Z ( x:t. ) , C ( h ) , is given by : C ( h ) ( 3 . 9 ) Unlike ( 3 . 5 ) , equation ( 3 . 9 ) has no denominator because 1 / ( n- 1 ) = 1 when there are only two obs ervation points , X;�. and ( x:1. + h ) . Second-order sta t i onari ty exists i f the value of C ( h ) for each pair of property values Z ( x;�. ) and Z ( x:t. + h ) is _ the same , and independent of its position in the sampling region ; that i s , C ( h ) depends only on h ( Trangmar et a l . , 1 985 ) , and the variabil i ty of z is the same throughout the region ( Russo & Bresler , 1 98 1 ) . By implication, when h is zero , C ( h ) must be equivalent 3 4 to the variance of z , of ten denoted by C ( O ) . The ratio of the spatial covariance to the sample variance is cal led the spa tial a uto-correl a tion coeffi cien t , P (h ) given by : P ( h ) = C ( h ) I C ( O ) ( 3 . 1 0 ) Thus , under second order stationarity , the mean and variance do not vary . P ( h ) = 1 when h = 0 and the spatial covariance decreases as h increases and so P ( h ) becomes a useful geostatistical tool s ince a plot of P ( h ) against h will give an indication o f the size of h for which values o f z remain correlated , or are spa t i a l ly dependen t . The a ssumption of s econd-order s tationarity upon which P ( h ) and C ( h ) depend i s regarded by many geostatist icians as too s trong for many spatial variables because of the tendency of estimates of the variance to vary w i t hout 1 im i t a s the s i ze of the area under investigation is extended ( O l i ver , 1 9 87 ) . As an al ternative to assum ing second-order stationari ty , the in trinsi c hypo th esi s of regi onal i zed vari abl e theory may be used . This assumes that equation ( 3 . 4 ) holds and that for a given value of h , the di f ference between Z ( x"- ) and Z ( x"- - + h ) has a f inite vari ance which i s independent of x "- , the position of the sample ( Webster , 1 98 5 ) : VAR [ Z ( x"- ) - Z ( x"- + h ) ] = E { [ Z ( x"- ) - Z ( x"- + h ) ] 2 } ( 3 . 1 1 ) = 2 ¥ ( h ) where ¥ i s the semi - variance . Implicit in the assumptions underlying equations ( 3 . 4 ) and ( 3 . 1 1 ) is that the soi l property fol lows the following model of variation : ( 3 . 1 2 ) where � � i s the mean value of Z in a region , v , and � ( x"- ) is a spatially dependent random component with zero mean , and a variance def ined by : 3 5 VAR [ � ( x� ) - � ( x� + h ) ] = E { [ � ( x� ) - � ( x� + h ) ] 2 } ( 3 . 1 3 ) = 2 ¥ ( h ) Thus , under the constraints o f the intrinsic hypothesis , variables need only be l o ca l l y sta t i onary . It wi ll be assumed for the rest o f the analysis of the 25 pH data that local stationarity applies . iv . The variogram A The semi-variance , ¥ ( h ) , is estimated by ¥ ( h ) for each value o f h where ( Webster , 1 985 ) : A ¥ ( h ) = 2: { Z ( x� ) - Z ( x� + h ) } 2 ( 3 . 1 4 ) 2m ( h ) 1\ ¥ (h ) is equivalent· to hal f the sum of the squared di f ference between pairs of values of Z ( x� ) and Z ( x� + h) averaged according to the number of pairs , m , at each value of the lag h . A. A plot of ¥ ( h ) against h for a range of separation distances i s the semi - vari ogram, which for simplicity will henceforth be cal led the vari ogram . The variogram represents the average rate of change of a property with distance ( Ol iver , 1 987 ) . Figure 3 . 3 shows the experimental variogram for the 25 pH data from the transect . Although there is much fluctuation in A A the value of ¥ ( h ) , i t can be seen that the trend i s for ¥ ( h ) to increase as the lag increases ; i . e . samples closer together have a lower semi variance than those farther apart , such that the var iance of the property is s aid to be spatially dependent . Figure 3 . 3 also shows how the number of pairs of points decl ines with increasing lag . From Figure 3 . 2 and the prel iminary analysis described in " section ( i i ) , it would seem likely that the values of ¥ ( h ) at large lags have low precision compared with those at small lags . Ol iver ( 1 987 ) noted that the precision of the variogram depends on the effective degrees o f 0. 1 6 0. 1 4 0. 1 2 I\ 0. 1 y (h) 0.08 0 .06 0 .04 0 .02 24 1 8 No: of 1 2 pa1rs 6 0 0 5 1 0 1 5 No. of units of lag 20 25 B- Experimental variogram Sample variance Figure 3 . 3 Experimental variogram for the 2 5 pH data , and the number of pairs of points separated by each lag 3 7 freedom a t each lag, which are a function o f the number o f pairs at each po int , and a l s o on the s ampl ing interva l , and degree o f s pa t i al variation . The dependence o f variogram precision on the ef fective degrees " of freedom is demonstrated by Figure 3 . 4 , which shows the change in ¥ ( h ) as the number o f pairs used to calculate i t increases , From Figure 3 . 3 i t can be concluded that as the distance separating samples increases , so ,.. does the value of ¥ ( h ) ; i . e . the variance between close samples i s less than the variance between points far apart . This , as has already been " suggested , could be due to the decreased precision of ¥ ( h ) at large h , so 1\ there may be a double effect ; increased ¥ ( h ) due to increased h ( Figure " 3 . 3 ) , and increased ¥ ( h ) due to a decrease in m ( Figure 3 . 4 ) . The former is the spa tial effect and one could reasonably conclude that the value o f 1\ h at which ¥ ( h ) ceases to increase , called the range, marks the l imit o f spa t i a l dependence , the f i nd ing o f which in many cases may be the " obj ective o f the geostatistical analysis . The value of ¥ ( h ) at this and greater values of h is known as the si l l . The s ignif icance of both s ill and range i s discussed more ful ly in section ( vi ) . Nevertheles s , to f ind the s i l l and range , a model must be f itted to the experimental variogram A so that its value can be interpolated . However , the increase in ¥ ( h ) due a possible lack of precision at large values of h presents a problem " 1\ since clearly , ¥ ( h ) at large h should carry less weight than ¥ ( h ) at small h . Thus , to account for the di f ferent value of m at diff erent values of h, weighted least squares must be used for the f i tt ing of variogram models ( the types of which are outl ined below ) . This point i s discus sed extens ive ly by K i tanidi s ( 1 983 ) , Armstrong ( 1 98 4 ) , Cressie ( 1 98 5 ) , and McBratney and Webster ( 1 986 ) . Figure 3 . 5 shows the experimental variogram for pH along the transect with two l inear models f i t ted - one , by weighted , and the other by ordinary least squares . As the models are linear , the range cannot be A identif ied s ince no point i s reached where ¥ ( h ) ceases to increase with increasing h . Given that the slope of the variogram is a measure of the degree or intensity of spatial dependence ( Ol iver , 1 98 7 ) , it c'an be seen that weighted and non-weighted models dif fer markedly ; the inte�sity of spatial dependence is shown by the weighted model to be much less than is indicated by the model fitted by ordinary least squares . The weighted variogram model takes the form : 0.5 0.45 0.4 0.35 0 .3 1\ y (h) 0 .25 0.2 0 . 1 5 0 . 1 0 .05 0 Figure 3 . 4 0 5 1 0 1 5 20 25 No. of pairs ,. Change in the value of Y ( h ) as the number of pairs of data points used to calculate it increases . Here h = 1 lag uni t ; the maximum number of pairs separated by this value of h is 2 4 . 0 . 1 6 0. 1 4 0. 1 2 0 . 1 1\ y (h) 0 .08 0 .06 0.04 0.02 0 5 0 ---- --o ---- 0 1 0 1 5 No. of units of lag o o 0 -- -- -- 0 20 25 0 Experimental variogram - - Sample variance - -Ordinary least squares Weighted least squares Figure 3 . 5 Experimental variogra� for soil pH with l inear models fi tted by ordinary and weighted least squares optimization 4 0 1\ ¥ ( h ) = 0 . 0 4 7 0 + 0 . 00 0 6 h ( 3 . 1 8 ) By def inition , ¥ ( h ) at h = 0 must be zero ( Webster , 1 985 ) . However , as can be seen in Figure 3 . 5 , the fitted model used here to approximate the sample semi-variance does not pass through the origin . Thi s fact that the A f itted variogram intersects the ordinate at a value of ¥ ( h ) greater than zero i s one of the more important aspects of the experimental variogram . " The value of ¥ ( h ) at this point , denoted by Ca is known as the nugget vari ance , and corresponds to either unexplained error or variabil ity of z which i s undetected at the scale of sampling , or both . In Figure 3 . 5 , the value of Co ( 0 . 0 4 7 0 ) i s not much less than either 52 ( 0 . 0 5 1 7 ) or the " value of ¥ ( h ) at h = 2 5 units of lag ( 0 . 0620 ) , and thus , this variogram i s s a id to have a h igh nugget variance ; i . e . the degree of spatial dependence is low . This point i s discussed in more detail in sect ion ( vi ) . v . Spatial variation in two dimensions S ince a transect will only show variation in one dimension , data sampled in the same area in two dimensions might be expected to show di fferent spatial varia tion, and it is therefore use ful to repeat the analysis outl ined above on data sampled on a grid . Figure 3 . 6 shows the spatial arrangement of a data set compris ing 1 2 1 pH measurements made on soil samples taken from the same area as the transect , this time on an 1 1 x 1 1 sampling grid with a grid spacing of 2 . 5 m . The 1 2 1 data points were found to be normally distributed ( Figure 3 . 7 ; R2 = 0 . 9 1 , p < 0 . 1 % ) . The mean , � ' was 4 . 92 and the sample variance , s2 , was 0 . 0 3 3 5 . Instead of analysing the data one point at a t ime as for the transect ( F igure 3 . 2 ) , th<:! data were ana lysed i n two dimens ions by s tarting in the top left corner and moving in stages a long both the north - so u th and ea s t - wes t axes s imultaneously , one lag at · a time . Throughout thi s analysis , iso tropy was assumed , that is , i t was assumed that the spatial variation ( i f any ) is the same in both the north-south - - and east-west directions ( and in a l l other directions ) . When the data are i sotropic the lag becomes a scalar ( Ol iver & Webster , 1 9 87 ) , and the data for a l l directions may be grouped , which is what has been done here . A B c D E F G H I J K a 5 . 38 4 . 5 3 4 . 9 5 4 . 90 4 . 7 5 4 . 80 4 . 7 5 4 . 80 5 . 0 0 5 . 1 8 4 . 98 b 4 . 38 4 . 68 4 . 8 0 4 . 98 4 . 65 4 . 90 4 . 80 4 . 85 5 . 0 0 4 . 9 5 4 . 9 5 c 5 . 00 4 . 9 3 4 . 85 4 . 68 4 . 68 4 . 7 0 4 . 98 4 . 80 5 . 0 3 4 . 8 3 4 . 9 5 d 5 . 1 0 4 . 80 4 . 9 0 4 . 53 4 . 8 0 4 . 7 3 4 . 9 3 4 . 9 0 4 . 9 8 4 . 63 4 . 58 e 4 . 65 4 . 88 4 . 7 0 4 . 85 4 . 9 0 4 . 9 5 4 . 9 3 5 . 05 5 . 0 5 4 . 7 8 4 . 7 0 f 5 . 2 0 4 . 88 5 . 0 8 5 . 0 5 4 . 78 4 . 9 3 4 . 9 0 4 . 9 0 5 . 20 5 . 3 0 5 . 08 g 4 . 7 8 4 . 9 0 4 . 83 5 . 20 4 . 80 5 . 2 0 5 . 1 0 4 . 7 0 5 . 1 3 5 . 1 3 5 . 28 h 5 . 38 4 . 9 3 4 . 7 5 5 . 1 0 4 . 98 5 . 2 0 5 . 0 5 4 . 9 3 5 . 0 5 5 . 1 0 5 . 3 5 i 4 . 9 5 5 . 0 5 5 . 0 3 5 . 0 5 5 . 0 3 4 . 9 3 4 . 9 3 4 . 9 5 4 . 75 4 . 78 5 . 0 5 j 4 . 78 4 . 8 5 4 . 9 0 5 . 0 0 5 . 0 5 5 . 0 5 4 . 83 4 . 80 5 . 1 3 5 . 0 5 5 . 20 k 4 . 7 0 4 . 8 5 4 . 7 8 4 . 78 4 . 98 5 . 0 5 4 . 7 0 4 . 7 0 4 . 70 5 . 00 4 . 93 Figure 3 . 6 1 2 1 equally spaced pH data . ( N . B . The ak and AK axes were of equal length in the f ield sampling grid . ) Normalised frequency 2 .5 2 1 .5 1 0 .5 0 0 +-�-=��----�-----+------�-----+----� 4.2 4.4 4.6 4.8 pH 5 Figure 3 . 7 Distribution of the 12 1 pH data shown in Figure 3 . 6 5.2 5.4 4 3 The data were assessed for changing mean and variance as the number of samples increased w i th increas ing lag, th is time in a progression dependent on the square of the lag ; 1 , 4 , 9 , 1 6 , 25 . . . . 1 2 1 , rather than in a s imple ari thmetic progress ion as was the case with the transect . As F i gure 3 . 8 shows , the precis ion o f the mean again increased with increasing sample number . Figure 3 . 9 shows the change in . the number of pairs of points , m, with increasing lag, h . As in Figure 3 . 3 , m decreases markedly as the lag increases . However , the degree . of precision is seen to be much greater at high lags in the case of the grid compared to the transect ( Figure 3 . 1 0 ) . In the case of the transect , there were only 2 4 pairs of points a t the smallest lag . With the grid, there are 22 pairs a t the largest , and 2 2 0 pairs a t the smal lest lag . Obviously , a much greater sampling ef fort is needed for 1 2 1 samples as compared to 2 5 , but these two figures i l lustrate the value of sampl ing in two dimensions ; in the case o f a transect -based sampl ing strategy , this could be done by sampling along two intersecting transects . Assuming isotropy , this would have the effect of doubl ing the precision of a variogram based on a s ingle transect since there would be twice as many pairs of observations . i . e . samples would have to be taken from 49 points instead of 2 5 ( the point at which the transects intersect would occur on both transects ) . The effect of changing the point of intersection on the variogram is not c lear and wi l l not be i nvestigated here , al though where the data are aniso tropi c the point of intersection and the angle of one transect to the other may be important . Figure 3 . 1 0 shows how the variogram changes as more samples are included by increasing the length of the side of the square grid to include more " l ags . In Figure 3 . 4 , ¥ ( h ) was shown to decrease as m increased when A samples separated by a s ingle unit of lag were considered . Here , ¥ ( h ) is shown to decrease as m at all lags increases , as is indicated by the development of the variogram from 2 lags ( 4 samples ) to 1 0 lags ( 1 2 1 ) samples . It is clear that the precision of the variogram for a l l 1 0 l ags is signif icantly greater than when h is less than 1 0 ; one infers , because ,... ¥ ( h ) decreases at intermediate lags as m for any one lag increases , that the precis ion of the variogram i s increasing . If there were spatial dependence , increas ing m through bringing in more data points might be 1\. expected to increase ¥ (h ) at intermediate lags . 5.4 5.2 Mean 5
4.8
4.6
0.2
0. 1 6
Variance 0· 1 2
0 .08
0.04
I
I
�e -o- -{)- - <:T - - o- - - e- - - -o- - - - e- - - - -o
0 �------+-------+-------�------�------�-------+
0 20 40 60 80 1 00 1 20
No. of samples
Figure 3 . 8 Change in the mean ( �) and variance ( s 2 ) with increasing sample number for
the set of 12 1 pH data
250
*... Length of grid side '
200 '*... 1 Jag unit ' +
G. '*... * 2 Jag units '
"R '
1 i0
'*... 3 Jag units '
'& '
No. of ' '*..... -e- 4 Jag units
pairs 'u.. ' X.. ' � -+ S lag units
1 00 ...... "0, 'X, ' 6 Jag units '*.....
......_
' -7( 7 Jag units '
�
50
"R ' 8 Jag units ' '
*..... "0, ' -e 9 lag units ' ......
0 '* 'X -« 1 0 lag units
0
0 1 2 3 4 5 6 7 8 9 1 0
No. of units of Jag
Figure 3 . 9 Change in the number of pairs of points separated by each lag as the length of the grid
side increases from 1 to 10 lag units
0. 1 2
Length of grid side
0 . 1 2 1ag units
-+ 3 1ag units
0.08 /
.....
* 4 1ag units
/
B S lag units 1\
y (h) 0.06 ....,_ 6 Jag units
-K 7 1ag units
0.04 -+- 8 lag units
"'* 9 Jag units
0.02 -e- 1 0 Jag units
0 1 2 3 4 5 6 7 8 9 1 0
No. of units of Jag
Figure 3 . 10 Development of an experimental variogra� as the number of samples used to estimate
it increases as the length of the grid s ide increases from 2 ( 4 s amples ) to 10 lag
units ( 12 1 samples )
4 7
vi . Other variogram models
Unlike the variogram in Figure 3 . 5 , the complete variogram for the grided
data ( Figure 3 . 1 0 ) has a curvil inear form . It i s therefore appropriate to
investigate some other possible variogram models .
Armstrong ( 1 984 ) noted that the eff icacy of geostatistics was essentially
dependent on the qual ity of the estimate obtained for the variogram .
Thus , getting the best fit for the variogram i s v'ry important , whether
it i s to be used for interpo lating va lues of z at unsampl ed s i tes
( kriging ) , or simply to f ind the values of h for which values of z are
related . It is not intended here to discuss kriging s ince it was not
necessary to use this technique in any of the work presented in this
thesis . Suf f ice to say that kriging involves minimizing the estimation
variance of the interpolated values of a property . One of its principle
advantages over other interpolation methods is that it gives a measure of
precision , but this is only good if the model for the spatial structure
( ie . the variogram ) is at least approximately correct ( Starks & Fang ,
1 982 ) . Kriging therefore rel ies heavily on the form and goodness of f it
of the variogram model , especially at small lags , since interpolation i s
invariably used to g ive inf ormat ion on the spaces between existing
sampling points .
The rationale which determines whether a particular model i s suitable to
describe a variogram is complicated and will not be dealt with here . It
i s d i s cussed extens ively by Armstrong and Jabin ( 1 98 1 ) , Christakos
( 1 9 84 ) , and W''Bratney & Webster ( 1 986 ) . Here it is enough to say that
l inear , spherical , and exponential models are permissible functions for
variograms and will describe most .
As indicated by equation ( 3 . 1 8 ) , the l inear model takes the general form :
¥ ( h ) = Co + kh for h > 0 ( 3 . 1 9 )
where Co i s the nugget variance , and k i s the s lope . Note that this model
has no s i l l , and therefore no range . In the case where k = 0 , the
variogram i s said to show a pure nugge t effect ( Webster , 1 985 ) ; that i s ,
there i s no spatial dependence a t the scale .of sampling .
4 8
The spherical model takes the form :
¥ ( h ) = Co + C { [ 3h/2a ] - [ (h/a ) 3 /2 ] } for 0 < h < a
( 3 . 2 0 )
¥ ( h ) = Co + C for h > a
and its tangent at h = 0 cuts the s i l l at 2a/3 . Here , a is the range , and
h , and Co are the lag , and nugget variance respectively . Just as Co
represents variabil ity which i s not detected at the scale of sampling , C
represents variabil ity which is detected at the scale of sa�pl ing; the
s i l l variance i s given by Co + C . In theory , the spherical model is
three-dimensional , yet Webster ( 1 98 5 ) noted that i t nearly always f its
exper imenta l results f rom s o i l s ampl ing be tter than one and two
dimensional analogs , such as the circular model ( which is not recommended
for describing variograms ) .
The exponential model takes the form :
¥ ( h ) = Co + C [ 1 - exp ( -h/r ) ] for h > 0 ( 3 . 2 1 )
Here , r i s a distance parameter which controls the spatial extent of the
function ( Webster , 1 985 ) . In the exponential mode l , ¥ ( h ) approaches the
s i l l asymptotically and there i s no such thing as a def inable finite
range . However , since the semi-variance must cease to increase beyond a
certain point , the range is taken to be equa l to 3r , where ¥ ( h ) is
approximately equal to Co + 0 . 9 5C ( Webster , 1 98 5 ) . Ol iver and Webster
( 1 9 87 ) have noted however , that taking the range as equal to 3r tends to
overestimate i t , and where this �ppears to be the case, the spherical
model will probably give a better fit because it curves more tightly .
Whichever model is chosen , it is pertinent to bear in mind that
" to serve us wel l , the model has to adequately portray the
behaviour of the measurements as they really are . I t is not
enough to represent how we wish the measurements had been ( but
were not ) . " ( Tukey ( 1 973 ) quoted by Armstrong ( 1 98 4 ) . )
49
According l y , w=Br a t ney and Webs t er ( 1 9 8 6 ) recommend the Aka i ke
Information Criterion ( AIC ) for determining which is the best model for
the experimental variogram . The A IC is calculated according to the
equation :
AIC n [ ln ( R ) ] + 2p ( 3 . 22 )
where n i s the number of observations ( or points on the variogram ) , p is
the number of estimated parameters ( s i l l , range etc . . . ) , and R is the
residual mean square of deviations from the f itted model . The model with
the smal lest AIC value is the best . Where the models being compared have
the same number of parameters ( spherical and exponential ) , there is no
need to calculate the AIC s ince a simple comparison of the residual sum
o f squares wi l l reveal which model has the best f i t .
The three v a r i ogram mode l s des er ibed above were f i t ted to the
experimental variogram for the 1 2 1 grided data and the AIC calculated for
each . The value of AIC was lowest when the spherical model was used ; this
model is shown fitted to the experimental variogram by weighted least
A
squares ·in Figure 3 . 1 1 . Here, ¥ ( h ) increases from a nugget variance of
0 . 0 1 83 to a s i l l of 0 . 0 5 09 at h = 3 . 9 6 lag units , the range . Thus , the pH
data show spatial dependence at sample separations less than 3 . 9 6 lag
units which is equivalent to 9 . 9 m .
In Figure 3 . 1 1 , and also in the other f igures showing the variograms for
both the transect and grided data , the sample variance has been plotted
as wel l as �he experimental variogram , and in Figure 3 . 1 1 i t i s seen to
be approximately equivalent to the s i l l . Trangmar et a l . ( 1 98 5 ) stated
that the s i l l corresponds to the maximum variabilty of z which i s
detected at the scale o f sampling . When the sample variance i s almost
pure nugge t , that is Co � s2 , the s i l l will be approximately equal to the
sample variance ( Webster , 1 985 ) . Generally the s i ll is expected to be
somewhat l arger than the sample variance ( Webster , 1 985 ) . One reason that
this might be expected to be so is that the f itted model represents the
bes t fit to a set of experimen tal data , although it might equally be
expected that the s i l l would under-estimate the sample variance for the
0.05
0 .045
0.04
0 .035
,...
y (h)
0.03
0 .025
0 .02
0 .01 5
0
*
*
* *
- - - -;r - - - --*- - - - - -
1 2 3
*
4 5 6
Lag (h)
*
Range = 3.96 lag units (9.9 m)
CO = 0.01 83; C = 0.0326
7 8 9 1 0
- - Sample variance
- Spherical model
Figure 3 . 11 Experimental variogram for the 1 2 1 pH data f itted with a spherical model by weighted
least squares optimization
5 1
same reason . Another is that the estimated variance s 2 must be smaller
when more samples are taken within a given area , especially when there i s
some spatially dependent variability . This point is discussed in some
detail by Webster ( 1 98 5 ) and will be returned to in Chapter 1 0 . For now ,
i t may be concluded that the spherical variogram shown in Figure 3 . 1 1 i s
a good model o f the experimental data as sampled on the grided design as
described .
vii . Conclusions
Using " aggie" s tatist ics ( Nielsen , 1 987 ) , the conclusions made as to the
1 2 1 pH data would have been conf ined to comments on the estimated mean
and sample variance , and in isolation these would tell very l i ttle . Of
course , if measurements of another property had also been made at each
xi , then correlations , comparisons and relationships between them could
have been investigated , but the often large error term may have occurred .
By performing a s patial anal ysis us ing geostatis tica l techniques as
detailed above , a large portion of this error term can be accounted for
in terms of the spatial effect . When the nugget - variance is low as a
proportion of the sample variance , then the spatial ef fect at the scale
o f sampling is large and may merit more intensive investigation . I f Co is
large in comparison with 52 , the error term may not be due to a high
degree of spatial dependence ; at least , no t at the sca le of sampl ing
used . This would either merit further spatial analysis at a di f ferent
sampling scale , or al ternatively may point to an inadequacy in the
analytical techniques used or to the use· of · an insufficient number of
samples . The other important information gained from a spatial analysis
i s the identi f ication o f the range ( i f present ) since i t indicates the
minimum sampl e separat ion for which samples are unrelated . In other
words , knowing the range to be equal to � , any further sampling should be
done with samples taken at intervals of h > � so as to avoid spatial
dependence in the data , and thus to e l im inate error due to sampling
design . This is the obj ect of the analysis o f spatial variabilit y in
nitri f ier activity which forms the bulk of the work described in this
thesis .
CHAPTER 4
EXPERIMENTAL METHODOLOGY
A . THE SHORT-TERM NITRIFICATION ASSAY
5 2
Studies of nitrification and the ef fects on i t of various soi l parameters
such as pH ( e . g . Frederick, 1 9 5 6 ; Aleem & Alexander , 1 9 6 0 ; Morri l l &
Dawson, 1 96 1 ; Darrah e t a l . , 1 986b) , or soil moisture and temperature
( e . g . S indhu & Cornfield, 1 9 67a ; Kowalenko & Cameron , 1 9 7 6 ; Addiscott ,
1 98 3 ; Nakos , 1 984 ; Macduff & Whi te, 1 98 5 ) have been undertaken f or a
number of years . Commonly these have involved either long-term perfus ion
experiments , pure culture studies of nitrifiers growing under laboratory
conditions , or incubation experiments of several days or weeks duration .
Thus , l ittle attention has been paid to monitoring the in si tu nitrifier
activity , either in i solation, or in response to changes in one or more
of the s o i l parameter s . The short-term n itri f i cation assay , SNA, of
Sarathchandra ( 1 9 78 ) and Schmidt and Belser ( 1 982 ) , as modif ied by Darrah
et a l . ( 1 986b ) , permits the study of in si tu nitri fier activi'ty without
the complication of microbial growth . Consequently , this technique , w ith
the modi f ications de scr ibed be low , was chosen as the basis of the
experimental work reported in this thesis .
The rate of nitrate production by nitrif iers growing under non-l imit ing
substrate conditions can be described by the equation ( Darrah et a l . ,
1 9 85b ) :
CIN03 = Jl\.1,. ... ,. • exp ( � ...... x t )
Cl t y
( 4 . 1 )
where CIN03/Ci t is the instantaneous rate of nitrate production , � ...... x is
the maximum specific growth rate ( h- 1 ) , m is the nitrif ier biomass ( � g
g- 1 soi l ) and Y is the growth yield constant ( � g biomass formed per � mol
of ammonium oxidized ) . When the incubation t ime t is short , m may be
assumed to be constant , and equation ( 4 . 1 ) simpl i f ies to ( Darrah e t a l . ,
1 9 86b ) :
oN03 = Jl1)max
a t Y
J J
( 4 . 2 )
This constant rate of nitrate production can be measured in the short
term nitrif ication assay . The rate of nitrate production in a s teady
state population should depend on the size of the nitrifier population
and t he phys i o l og i ca l a ct iv i ty o f the organ i sm s mak ing u p that
population . Thus , the measured nitri fication rate in a short-term assay
reflects these features of the population and w i l l be referred to by the
a l l-encompasing term , ni trifier acti vi ty; the SNA value is an index of
nitri fier activity .
S ince the test of no measurable growth i n a n itr i f ier population i s
l inearity of nitri f i cat ion , the succes sful use of equation ( 4 . 2 ) in
l aboratory incubation experiments depends upon the selection o f a t ime
interval for incubation such that nitrif ication proceeds linearly over
the who l e period t . I t a l so depends on the use of a non-limiting
incubation medium and substrate concentration . Thus , the SNA technique
has to be tailored to suit the particular soil under investigation .
i . Selection o f incubation medium for SNA analyses
The observations of D arrah et a l . ( 1 98 7 a ) of the adverse effect on
nitrif iers of solutions of low osmotic potential suggest that incubation
media should have an ionic strength approximately equivalent to that of
soil solutions in the f ield . Edmeades et a l . ( 1 98 5 ) studied the chemical
composition and ionic strength of a range of New Zealand topsai l s under
grassland and found that ionic strengths ranged from 0 . 0 0 3 - 0 . 0 1 6 mol dm- 3
with a mean value of 0 . 00 5 . Dol l ing and Ritchie ( 1 985 ) found that the
ionic strengths of soil solutions from 2 0 soils from Western Australia
had very similar values . They also noted the marked ef fect of dif fering
ionic s trength on the measurement of soi l pH . pH measurements made in
s o l u t i on s w it h a n i on i c s t reng th of 0 . 0 0 5 d i f fered leas t f rom
measurements made at the ionic strength of so il solutions at f ield
capaci ty , whi l s t the d i f ferences that o ccurred in comparisons with
5 4
dist i l led water or CaCl2 a t an ionic strength o f 0 . 0 3 ( 0 . 0 1 M ) were much
greater ( ;:: 0 . 4 pH units ) . In view of the dependence ·of nitrif ication on
pH ( Darrah et al . , 1 9 86b ) and the fact that the relationship between pH
and nitrif ier activity was to be a maj or area of interest in this study ,
it was clear that all SNA measurements had to be made in solutions of
ionic s trength close to that of the soil solution in the field . The mean
ionic s trength of the topsoi l of Tokomaru s i lt loam was measured by
Edmeades e t al . ( 1 985 ) as 5 . 4 x 1 0- 3 mol dm-3 ( wi th a range of 2 . 1 - 1 1 . 3 x
1 0- 3 mol dm-3 ) . On this basis , it was decided that 0 . 0 05 M KCl would be
sui t ab le for SNA ana lyses and as a medium for all the experiments
reported here ( unless otherwise stated ) ; the dif ference in ionic strength
between 0 . 005 M KCl and the expected extremes in ionic s trength of f ield
soi l solutions was assumed to be suf f iciently small to cause nei ther
dispersion of the soil nor inhibition of the nitrif iers .
ii . Linearity of nitrification rate in the Tokomaru Silt Loam
Aleem and Alexander ( 1 960 ) found that the minimum generation time for
Ni trobacter agi l i s was about 7 hours , whilst Sarathchandra ( 1 9 78 ) found
no s ignif icant change in the most probable number of soil nitrif iers
during 1 7 hours incubation . Accordingly , Sarathchandra ( 1 978 ) and Steele
et a l . ( 1 9 80 ) sampled incuba ting med ia a f ter and 1 7 hours and
cal cul ated SNA values as the dif ference between the amount . o f N03 -N
produced between these times per g soil per hour . Darrah et al . ( 1 986b)
cal cul ated SNA values in the same way but used shorter incubations ,
sampling a fter 1 and 8 hours fol lowing the addition of NH4-N substrate .
In v iew o f this range of incubat ion t imes for SNA measurements , an
e x p e r i men t w a s c onduc t ed t o i nv e s t i g a t e the 1 ineari ty of the
nitrif ication rate in the Tokomaru silt loam .
5 5
Methods and Materials
A bulk soil sample ( approx . 5 kg ) was dug from the 3 -9 cm depth range of
a randomly selected s i te in f ield No . 6 ( see Section B , below ) . The soil
was sieved ( < 2 mm ) , thoroughly mixed , and a 200 g subsample was placed in
a Buchner funnel f itted with a Whatman No . 1 f i lter paper , and leached
overnight with 1 dm3 0 . 005 M KCl to remove any nitrate present . At the
end of leaching , excess moisture was removed from the ·soil by suction
f i ltration for 9 0 minutes , after which 3 8 replicate 5 g samples ( oven-dry
equivalent - determined by oven-drying overnight at 1 05 °C ) were placed
into 5 0 cm3 incubation tubes containing 20 cm3 0 . 00 5 M KCl with 0 . 3 % �;v
agar . In l ater exper iments the d i lute agar suspension a l lowed the
incubating media to be sub-sampled after 1 hour without affecting the
soil : solut ion rat io for the rest of the i ncubat ion ( Darrah et al . ,
1 98 7a ) .
1 0 cm3 0 . 0 1 M ( NH4 ) 2S04 was added to each tube and the tubes were shaken
at 2 2 oc in an enclosed end-over-end shaker f itted with a thermostat .
( The choice of 2 2 oc as a suitable incubation temperature was entirely
arbitrary , and was governed by the fact that s ince the temperature inside
the shaker was mai nta ined by two 1 00 wat t l ight bulbs , 2 2 oc was a
temperature that was easily maintained at a constant , at all t imes of the
year . ) At hourly intervals up to 1 9 hours after the start of incubation ,
two tubes were removed , shaken and dupl icate 5 cm3 samples of suspension
were quickly taken by pipette from each tube . These were centrifuged at
3 0 0 0 r . p . m for 1 0 minutes and the supernatant frozen and s tored . The
solutions were later analysed for N03 -N on a Technicon Autoanalyser
fol lowing the method of Downes ( 1 9 78 ) .
Results and Discussion
Figure 4 . 1 shows the amount of N03 -N produced in each incubation plotted
as a function of the t ime of incubation . By expressing the data on an
hourly basis , i t was clear that the nitrif ication rate over the f irst
hour was significantly higher than it was during the remaining eighteen
hours of the incubation . Thus , the assumption of l inear nitrif ication did
1 .6
1 .4 0 �
1 .2 ��
/
1 o�'S
pmol §
N03-N 0.8 � z o
g-1 /
0 .6 A5
�6
0
0.4
0 .2
0
0 2 4 6 8 1 0 1 2 1 4 1 6 1 8 20
Time (hours) after addition of NH4-N
Figure 4 . 1 N itrification rate in the Toko�aru silt loam measured at 2 2 °C for
1 9 hours after the addition of a��oniu� substrate
5 7
not s eem to be a good one . In f act the data were best-f i tted by an
upwards curvil inear model ( Figure 4 . 1 , dotted l ine ; R2 = 0 . 98 , p< 0 . 1 % )
whi ch i s what would be e xpec ted for a growing population . _ It was
therefore concluded that the incubation time selected would have to be a
compromise which satis f ied three basic requirements ; ( a ) that ni trate
production should be approximately l inear with time ; ( b ) the incubation
t ime would have to be long enough to al low for production of a measurable
increase in No3 - concentration ; and ( c ) not too long that s ignif icant
popu l a t i on gr owth occurred , or the experi ment be came ph ys ical l y
impossible t o do . A l inear model was f itted to the data between 1 and 8
hours ( Darrah e t al . , 1 98 6b ) and s ince this gave a good f i t ( Figure 4 . 1
s o l id l ine ; R2 = 0 . 8 2 , p < 0 . 1 % ) these times seemed sui table for SNA
measurements on the Tokomaru silt l oam .
As indicated above , the time taken to complete an SNA analysis was an
important consideration . I t was apparent that the maximum number · of tubes
that could be dealt with at a time without incurring a t ime error whilst
sampl ing the suspensions , was between 1 5 and 20 . ( In fact with practice ,
1 5 tubes could be sampled in 2 minutes . i . e . the t ime o f sampl ing given
as 1 or 8 hours is accurate to ± 3 % after 1 hour and ± 0 . 4 % after 8
hours . ) By starting the Buchner suction at 6 . 30 am ( and allowing for a
period of 5 hours for the samples to equil ibrate fol lowing the addition
of acid or a l ka l i - s ee Chapter 7 and Darrah e t a l . ( 1 9 8 6b ) ) , and
separating the tubes into four groups of 1 5 with a stagger of half an
hour between the f irst and second and the third and fourth groups , and 4 5
minutes between the second and third , i t was possible to complete 6 0 SNA
analyses by . midnight i f an incubation of 8 hours was used . Accordingly ,
in a l l SNA anal yses reported here , the incubating media were sampled
after 1 and 8 hours , and the SNA value calculated as the difference
between N03 -N produced at these t imes per g soil per hour .
5 8
iii . Selection o f ammonium substrate concentration for SNA analyses
Macduff and Whi te ( 1 985 ) measured nitrif ication rates over a range of
so i l mo i s ture cont ents and incubation t emperatures , and found that
irrespective of temperature, nitrification was l imited by the supply of
NH4 -N . Gilmour ( 1 9 8 4 ) predicted that nitri f ication rates fol lowing zero
order kinetics w i l l increase as the i n i t i a l c.<:mcentration of NH4 ·
increases according t o the equation :
NR = k )I NXt ( 4 . 3 )
where NR is the absolute nitri fication rate , k is the rate constant , and
NXt is the initial NH4 • concentration at the start of the t ime period t .
Mo l ina ( 1 9 8 5 ) a s serted that n itr i f icat ion proceeded from pulses of
ammonium oxidation generated by microbial c lusters . The size of the pulse
must be subj ect to a negative feedback system , however , because if large
app l icat ions o f ammonium are suppl ied , the e f f ect of l ow osmotic
poten t i a l w i l l inhibit ni tri f i ca t i on ( Darrah et al . , 1 98 7 a ) . Thus
equation ( 4 . 3 ) must be treated with circumspection because i t suggests
that NR will continue to increase l inearly with increasing initial NH 4
concentration even at very high concentrations . Furthermore , Clay e t al .
( 1 985 ) noted that i f clusters of NH4 · oxidizers were to generate a pulse
of N02- large enough , their microni che may be acidi fied to toxic levels
and nitrificat ion would be reduced ( Chapter 7 ) . Neverthe less equation
( 4 . 3 ) does indicate that nitri fication rates can be limited by inadequate
NH 4 -N substrate concentrat ions . Thus , the concentrat ion of NH 4 -N
subs trate suppl ied i n SNA incubations must be such that i t is non
l imiting in the sense that it is in excess , but not so much so as to
generate conditions which are toxic to the nitrif iers .
The form in which the ammonium was to be supplied for SNA measurements
was also an important considerat ion , the obvious choice being between
NH4Cl and ( NH4 ) ;so4 since these are readily available forms of ammonium .
Darrah et al . ( 1 9 8 5 a ) monitored the response of soil ni tr i f iers to
additions of both NH4Cl and (NH4 ) 2S04 . Additions of more than 7 . 3 �moles
N g_ , soil as NH4Cl were found to inhibit nitrif ication , but . a s imi lar
e ffect was not found with (NH 4 ) 2S04 sugges ting that the chloride ion
59
rather than osmotic potential was the cause of the inhibition . Further
work ( Darrah et al . , 1 987a ) demonstrated that the inhibitory ef fect of
the cl - ion was disproportionate to its contribut ion to the . o smotic
pot en t i a l of the s o i 1 s o lu t i on . i . e . C l - i s t ox ic to nitri fiers .
Accordingly , ( NH4 ) 2S04 was chosen as the substrate to be used for SNA
mea su r ement s , bu t an experi men t was requ ired t o e stabl i sh what
concentration should be used, s ince Darrah et al . ( 1 985a ; 1 987a ) had
shown that this salt was also inhibitory to nitr i f ier activity in high
concentrations .
Methods and Materials
A bulk soil sample was col lected as for the l inearity experiment ( Section
i i . above ) and leached overnight with 0 . 0 05 M KC! . After the removal of
exce s s m o i stur e , 6 0 incuba t i ons were set up f o l l owing the method
described above , except that the tubes were spl it into 6 groups and
( NH4 ) 2S04 substrate added as 1 0 cm3 of 0 . 0 0 1 , 0 . 0 0 5 , 0 . 0 08 , 0 . 0 1 0 , 0 . 0 1 5
or 0 . 0 2 0 M ( 1 0 repl ica tes each ) . The incuba tions were carried out as
described , with sampling of the suspension after 1 and 8 hours .
Result s and D iscussion
The mean SNA values for each concentration of ( NH4 ) 2 S04 added to the
incubations are shown in Table 4 . 1 . An analysis of variance showed that
there was no signi f icant di f ference in SNA value ( p < 0 . 1 % ) over the range
of NH4 -N concentrations tested . i . e . there was either no inhibition of
nitr i f ier activity in any treatment , or the amount of inhibition was the
same for e ach treatment . Direct comparison between thes e results and
those o f Darrah et al . ( 1 987a ) i s made diff icult by the fact that ( a ) in
this experiment there was no means of measuring the osmotic potential ( an
osmometer was not available ) ; ( b ) the ionic strengths of the incubat ion
media were quite dif ferent ( 0 . 0 1 M CaCl2 compared to 0 . 00 5 M KC! ) ; and
most importantly ( c ) the ionic s trength of the soil solution in the f ield
in the soil studied by Darrah et al . ( 1 987a ) was probably different from
that in the Tokomaru s i l t loam , wi th the consequence that the soil
Table 4 . 1 Effect of ( NH4 ) 2S04 substrate concentration on SNA value
Concentration of
( NH4 ) 2S04
0 . 00 1 M
0 . 005 M
0 . 0 08 M
0 . 0 1 0 M
0 . 0 1 5 M
0 . 020 M
Mean SNA
( IJ mol N03 -N
g- 1 h_ , )
0 . 0 1 6
0 . 0 1 5
0 . 0 1 5
0 . 0 1 5
0 . 0 1 3
0 . 0 1 5
Standard
e:r.ror
0 . 00 1 0
0 . 0008
O . OO J 1
0 . 00 1 5
0 . 00 1 0
0 . 00 1 7
6 1
ni t r i f ier response to changes in osmotic potential might be quite
d i f ferent between the two soi ls , certainly in terms of i ts magnitude .
Nevertheless , i f one assumes that the dif ferences between the incubating
media were insignif icant in terms of their ef fect on the overal l osmotic
potential in the two experiments , s imple comparisons can be made .
Osmotic pressures were calculated for each substrate concentration using
the equation ( Hi l lel , 1 9 80 ) :
r = MRT ( 4 . 4 )
where r is the osmotic pressure ( Pa ; 1 bar = 1 0 5 Pa ) , M is the total
mol ar concentration of solutes ( mol m- 3 ) , T is the temperature ( degrees
Kelvin ) and R is the gas constant ( 8 . 3 1 4 J K_ , mol - ' ) . Using equation
( 4 . 4 ) , it was apparent that the osmotic potential of the substrates used
was high ( osmotic potential = - osmotic pressure ) in relation to those
of Darrah et al . ( 1 987a ) , and therefore one was led to conclude that
there was no inhibition of nitrif ication in any of these incubations . In
view o f thi s , and mindful that no account had so far been taken of the
pos s ible seasonal f l uc tuat ion in SNA value ( see Chapter 7 ) and the
consequent possibilty of much higher values at other times of the year
than those measured here , 0 . 0 1 M ( NH4 ) 2S04 was chosen as the substrate to
be used in a l l future SNA analyses . S ince the SNA value for 0 . 0 1 M
( NH 4 ) 2 S04 was 0 . 0 1 5 ± 0 . 00 1 5 � mol N g- ' h_ , indications were that this
concentration of NH4- was wel l in excess of the nitrif ier requirements ,
but not too high to cause toxicity problems .
iv . Analysi s o f exchangeable ammonium
In several o f the experiments described in the fol lowing chapters ,
exchangeable ammonium was analysed . This was a particularly important
ana l ys i s in the spatial vari abil ity experiment ( Chapter 6 ) s ince one
po s s ib i l ty w a s t ha t v a r i ab i l i ty in n i tr i f ie r act ivi ty f o l lowed
variability in Ex-NH4- ·
Ex-NH4 - was analysed by the fol lowing method ( A . D . A . S . , 1 98 1 ) :
62
6 g soil ( oven-dry equivalent , s ieved < 2 mm ) was weighed into 50 cm3
plastic tubes ( the same as those used for SNA measurements ) . To these , 3 0
cm3 2 M KCl was added and the tubes were shaken i n an end-over-end shaker
for 2 hours . The suspensions were then f i ltered through Whatman No . 3 2
f i l ter paper , and the f i l trate frozen and stored . This was l ater analysed
for NH 4 -N on a Technicon autoanalys er following the standard method
( Technicon Users Manual , 1 9 7 6 ) .
B . FIELD SAMPLING
i . S ite Details
Unless otherwise indicated , a l l the experimental work reported here was
carried out using soi l sampled from the Massey University No . 4 Dairy
Farm , specifically from two adj acent f ields in the Soil Science Research
Area - No . 6 , which had no previous history of ferti l izer trials or other
experimental work , was used for the bulk of the study , and No . 2 , a
f ormer lime tria l , whose history is detai led in Chapter 7 , was used for
the pH related work . Both f ields had been under a ryegrass-white clover
pa s tur e for severa l years and for the duration of thi s study were
periodically grazed by sheep ( 3 - 4 days each grazing ) at approximately 1 40
s tock units ha- ' ; one stock unit is equivalent to a 5 5 kg ewe ( weight at
mating ) which consumes suf ficient dry matter to produce one weaned lamb
per year ( Cornforth & Sinclair , 1 984 ) .
The soil at this site , the Tokomaru silt loam ( Cowie , 1 9 7 4 ) i s classi f ied
as a Yellow Grey Earth ( Taylor & Pohlen , 1 968 ) or Typic fragiaqualf ( So i l
Survey Staf f , 1 97 4 ) . I t i s a poorly drained so i l of low nutrient status
and was not recommended by Cowie ( 1 97 4 ) for cropping or horticultural
use . He recommended regular dressings of phosphate , l ime and potash to
achi eve opt imum pasture growth . Accordingl y , land on this soil is
typical ly used for town dairy supply ( serving Palmerston North ) and the
f a ttening of sheep , a l though product ion can be l imited by the wet
conditions in winter and spring , and by the drying out of the soil in
summer .
63
Mean annual rainfal l at the s ite , which is approximately 7 5 m above sea
level , is 9 9 5 mm and mean monthly temperatures range from 8 oc in. Jul y to
1 7 oc in February ( New Zealand Meteorological Service ;· Figure 4 . 2 ) .
ii . Soil Sampling
Soil samples for a l l experiments except the pH work ( Chapter 7 ) were
taken in sections using a core auger 3 cm deep with a diameter of 5 cm .
For a l l SNA measurements ( see below ) , the top 3 cm layer was discarded to
minimise any inhibitory ef fects that grass roots may have on the rate of
nitrif ication ( Mol ina & Rovira , 1 9 6 4 ; Neal , 1 9 69 ; Moore & Waid , 1 9 7 1 ) ,
and the 3 - 9 cm layer retained for analysis . The rationale behind using
this depth range for experimental work is more ful ly explained in Chapter
5 . On a l l sampling occasions , samples were sieved ( < 2 mm ) as soon as
possible a fter sampling and stored in sealed plastic bags at 3 oc ( see
Section C below ) .
Prior to soil sampling , there was always a period of three weeks during
which there was no grazing . This was done in an attempt to minimise
grazing effects such as ho tspo ts of high nitrate concentration caused by
the urine and excreta of grazing animals ( Ryden e t al . , 1 98 4 ; Bal l &
Ryden, 1 9 8 4 ; White , 1 98 4 ) .
iii . Correlation between moisture contents of sieved and rinsieved soil
As out l ined in section A ( above ) , the amount of s oi l required for a
s ingle SNA analysis was approximately 5 g ( oven-dry equivalent ) of s ieved
soi l . Each soil sample was analysed in duplicate for nitri fer activity ,
and in a ddi t i o n , t wo f urther sub - samples o f 6 g e ach ( oven - dry
equivalent ) were needed for analysis of exchangeable ammonium , and a
further 1 0 g needed for measurement ( in duplicate ) o f the soil moisture
content so that results could be calculated on a per g dry soil basis .
Overall , approximately 3 5 g sieved soil was needed for analysis . S ince
the purpose of this work was to study field n itrif ier activity , e ach
a. Precipitation
200
1 50 -e Evaporation
mm 1 00
-w- Rainfall
50
0 �--P---P---+---+---+---+---+---+---+---+-�
b. Temperature
20
1 6
OC 1 2
8
4 �--P---P---+---+---+---+---�--�--+---+-�
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
Figure 4 . 2 Mean monthly weather data ( 1 9 2 8 - 1 9 8 0 ) for the D . S . I . R Grass lands weather
station , Palrnerston North ( N . Z . Meteorological S ervice , 1 9 8 3 )
65
samp le had to be s ieved in the f i eld-moist s tate to ma intain field
conditions , and as a resul t , the amount of sieved soil gained from a 5 cm
diameter core from the 3-9 cm depth range was often not far in excess of
the amount needed for analysis , although this depended on the moisture
content of the soi l . Therefore in order to conserve soi l , i t was decided
to investigate the possibility of a relationship between the mois ture
content o f s ieved and unsieved so i l samples i n t he hope that field
moisture contents could be inferred from the s ieved moisture content .
This was parti cularl y important for the spatial vari abi l ity analyses
s ince the variability in moisture content was considered a possible cause
of variabil ity in nitrif ier activity .
Methods and Materials
Fifty soil samples were taken from sampling sites randomly arranged at
roughly 5 m intervals in field No . 6 at a range of depths to 24 cm using
the core auger . Sub-samples were assayed immedia tely for gravimetric
moisture content by oven drying overnight at 1 0 5 oc . The remainder of
each sample was s i eved ( < 2 mm ) and this too , was dried by the same
method .
Results and Discussion
The moi sture content of s ieved samples was plotted aga inst that of
unsieved samples and an equation fitted by l inear regression ( Figure 4 . 3 ;
R2 = 0 . 9 1 , p < 0 . 1 % ) . The equation took the form :
S = 0 . 0624 + 0 . 7 49 1 1 U ( 4 . 5 )
where S and U were the moisture contents of the s ieved and uns ieved
samples respectively .
In view of the good fit of equation ( 4 . 5 ) to the data over a wide range
of mois ture contents ·( 0 . 2-0 . 5 g g- , ) , this equat ion was used in all
subsequent experiments to calculate the field soil moisture content on
the basis. of the moisture content of the sieved soi l .
9g (gg-1 )
<2mm
SEfVED
0.45
0 .4
0.35 0 0
0.3 0
0.25
0 .2
0 . 1 5 +---+----+----t----'t---�1----t------f
0 . 1 5 0.2 0 .25 0.3 0.35 0.4
9g (gg-1 ) UNSEfVED
0.45 0 .5
Figure 4 . 3 Correlation between the gravimetric moisture contents of s ieved ( <2 �� )
and unsieved Tokomaru s i lt loam
C . STORAGE OF SAMPLES PRIOR TO SNA MEASUREMENTS
67
It was frequently necessary to col lect soi l samples several days before
laboratory measurements were made . Therefore the samples had to be
stored . The term s torage often encompasses drying, pulverization and
sieving ( Bartlett & James , 1 980 ) , and soils are often stored with l ittle
regard for the possible ef fects of storage on the soi l ' s properties . The
physical properties of soils are well known to be affected by drying ( and
wet t in g ) ; the contract ion o f pores i s .known to . occur and further
structural al terati ons may also arise . In extreme cases , such as in
Vertisols , the development of large cracks and fissures upon drying i s
l i ke l y , a nd w i l l be espec i a l l y s ignificant in larger , more massive
samples . In view of these known physical changes , i t appears unlikely
that the microbial and chemical characteristics of a soil would remain
unaltered from the field state during storage . Since the pur�ose of this
study was to investigate the activity of ni tri f iers in the f ield , the
method o f s ample s torage and pr eparat ion was obvi ous l y of gre a t
importance .
i . Effects o f drying and storage on mineral nitrogen
It has o ften been observed that variation in pretreatments , particularly
degree of drying , affects the amount of soil nitrogen mineral ized during
short periods , thereby complicating the use of the results as indices of
soil nitrogen availabi lity ( Stanford & Smith , 1 972 ) . Patten e t al . ( 1 980 )
s tudied the effects of drying and a ir-dry storage on the soi l ' s capacity
for denitrif ication under anaerobic incubation . Their results indicated
that the d ry ing o f s o i ls markedly increased the i r capac i ty for
deni tr i f icat ion , and this e f fect increased a s the drying temperature
increased . Partial dry ing and air-drying had a similar but slightly
sma l ler ef fect . Agarwal et a l . ( 1 9 7 1 ) found that t he · t emperature of
drying as wel l as the drying-rewetting cycles enhanced both nitrogen and
carbon mineralization in "practically all soils" studied . The N release
increased i f incubation followed drying . In all but one of their soi l s ,
air-drying caused a greater N release than heating at 6 0 oc regardless o f
68
the number of rewetting-dry ing cycles . Ross et al . ( 1 9 79 ) found that
sample s ieving led to a sl ight increase in all mineral N fractions as did
storage of the sieved samples for 24 hours at 4° or -20 oc . · Frye and
Hutchinson ( 1 98 1 ) f ound a pronounced increase in exchangeable NH 4 ... on
dry ing , this increase being greater a fter oven-drying . However , the
source of this NH4 ... was unknown although the effect was less in the
subsoi l suggesting that the NH4 "'" was released in the topsoi l . Thus ,
either humus or mi crobes kil led by drying ( or both ) were a possible
source of mineralizable N . These results agree with those of Soul ides and
Al l i son ( 1 9 6 1 ) , who f ound a s imi lar increase in available ammonium .
Gasser ( 1 9 6 1 ) found that drying led to an increase in both NH4 "'" and No3-
and found that on rewetting , most extra mineralization had occurred after
1 0 days and all by 42 days . He also found that the increase in nitrogen
mineralization became more marked with time of storage although a clear
trend was not e s tabl ished f o r 1 2 - 1 6 we eks be fore whi ch , v a lues
fluctuated . Munro and Mackay ( 1 9 6 4 ) found that incubation of soils at a
humidity of less than 85 % severely restric ted N03 - production . This
effect was due to drying as the soil moisture content at 85 % relative
humidity was hal f that a t a rel ative humidity of 1 0 0 % . However , on
rewetting the drier samples , the N03- production increased signi ficantly .
They also found that air-drying the soil from field capacity to wilting
point had little or no ef fect on No3- production , but further drying due
to a ir-dry storage caused a marked increase on rewetting .
Tests on C02 evolved during the experiments of Pat ten e t a l . ( 1 9 80 )
indicated that the increase in the soil ' s capacity for denitrification
was due to an increase in soil organic matter which was readily utilised
by denitrifying organisms . This was thought to be due to humus breakdown
although there i s good evidence ( Jenkinson , 1 9 6 6 ; Jenkinson & Powlson ,
1 97 6 a ; 1 976b ) to suggest that much of this increase - in organic matter is
in the form of dead organisms that were kil led off by drying . This is
supported by the results of Agarwal e t al . ( 1 9 7 1 ) , who with respect to
mineralization , proposed that in addition to microbial s timulat ion in
rewetted samples fol lowing drying , heat was directly responsible for the
amount of N and C released i n unincubated samples through chem ical
alteration of otherwise unavailable organic matter , and by the "kill ing
o f f " o f organisms . When incubat ion followed the drying and heating
69
treatment , the direct effect of heat together with increased microbial
act ivity and associated changes during incubation accounted for C and N
mineral ization .
Addiscott ( 1 983 ) noted that the net amount of N ammoni f ied in t ime t , Nt ,
measured as the increase in the sum of ammonium and nitrate , increased
approximately l inearly with t . However , this zero-order relationship was
found to be dependent to some extent on whether the soi l was pre-dried .
Many authors ( e . g . Stanford & Smith , 1 9 7 2 ; Tabor e t a l . , 1 985 ; Clay et
al . , 1 9 85 ) used soils that had been dried and sieved and onl y a very few
( e . g . Addi scott , 1 98 3 ) used soi ls in the f i eld state . It was thought
conceivable that the f lush o f mineral N caused by rewetting the dry soil
cou l d change a l i near re l a t i on sh ip wi th t i n t o an apparent .ft
relationsh ip, or even a f irst -order ( exp ( t ) ) relati onship ( Addiscott ,
1 98 3 ) .
Despite a l l this apparently conclusive work , it has been found that the
e f f e c t o f storing s o i l s on the ir subsequent abil ity to mineralise
n i t rogen depends on the indiv i dual so i l ( Harding & Ross , 1 9 6 4 ) .
Investigations of four soils showed the effect to be much more pronounced
in three but not so great in the fourth . These d i f ferences were related
to the moisture and carbon contents of the soils but i t was found that
for a g iven drying period , the amounts of carbon and nitrogen mineralised
were proportional to the carbon content of the soi l , whi le for a given
so i l , they were f ound to be a s i gn i f icant l inear function of the
logari thm o f the t ime the s o i l was in an a ir -dry s t a te prior to
moi stening .
i i . The e ffects o f drying and storage on soil biomass
From the above , i t is clear that the major effects o f soil drying and
s torage are governed by the e f fects of s torage on the microbial
popul a t ions s ince dead organi sms add to the pool of mineralizable
substrate . Thus , the mode and t ime of storage were especially important
to this study , given that it was the activity of the nitrifiers that was
under investigation .
7 0
Chao and Alexander ( 1 982 ) studied the inf luence o f drying on the survival
of Rhi zobi um . They found that the numbers of both R. mel i l oti and R .
pha seo l i f el l markedly a s the soils dried , but the ir abundance only
dec l ined s lowly in soi ls maintained in the air-dry state . Further , the
number of surviving cel ls increased i f the bacteria were added to sterile
soil and allowed to grow before desiccation was extensive . The' work of
stevenson ( 1 9 5 6 ) on the respiration of air-dried and fresh soils showed
that a higher level of metabolic activity is attained in air-dried soi ls
on remois tening than occurs in fresh soi l . The degree by which the
metabolic activity increased varied directly with the concentration o f
free amino acids and other nitrogenous materials released by the air
drying process . Salonius ( 1 9 83 ) allowed microbial populations from dried ,
remoistened and undried forest organic horizons to recolonise steri l ised
forest hor i z ons and conc luded that samples o f forest organic soi l
material designated for the study of microbially driven processes should
not be air-dried . White ( 1 9 6 4 ) reached a simi lar conclusion with regard
to the study of soil phosphate potential . He found that when air-dried
samples were used for measurements of phosphate potential , microbial
uptake of phosphate interfered with the resul ts if the samples were
shaken f or more than two hours . i . e . re-wett ing led to increased
microbial activity and thus uptake of P , this effect being significant
when samples were shaken for more than two hours .
Harding and Ross ( 1 9 6 4 ) added ammonium to dried stored soi ls and obtained
results suggesting that numbers of nitrifying organisms decreased after
six months of storage . Soul ides and Al l ison ( 1 9 6 1 ) found that drying was
more destructive to organisms than freezing and so the latter would seem
preferable if storage is required . Far more desirable however , would be
to use fresh soi l . Salonius ( 1 983 ) reported the work of Sneath ( 1 9 62 ) who
studied dormant populations in dry soil stored for up to three hundred
years and estimated that these soils would reach steri l ity in a thousand
years at ambient temperature and humidity . The d i f ference between s ix
months and a thousand years storage is obvious , but this hypothesis would
nevertheless seem both reasonable , and with clear implications .
7 1
Invest igations on the ef fec t of storage on soil biomass estimated by
biochemi cal techniques were carried out by Ross e t a l . ( 1 9& 0 ) . For
determination of biomass by C02 evolution from chloroform fumigated soils
incubated for f ixed periods , the differences in patterns of C02 evolution
between soils s tored for 2 8 and 5 6 days at 2 5° , 4° , and -20 oc were
negligible and not s igni ficantly different from those calculated from
individual ly determined incubation periods for each _ _ treatment and soi l .
However , biomass C va lues could change s igni f i cantly at all storage
temperatures but general ly least at -20 oc , the temperature of storage
which was bes t for maintaining ATP content s . Overa l l , no storage
temperature was satis factory for a l l indices of microbial biomass tested ,
but 4 oc was adequate for short periods . I t is clear therefore that
assays of fresh soil are preferable .
Mention has been made of the effect of drying on soil organic components
such as amino acids . B irch ( 1 9 58 ) showed that C and N mineralizat ion
occurred rapidly on rewetting dried soi l , the ef fect being greater for
oven-dried than for air-dried soil . He therefore concluded that �rying of
any kind leads to humus decomposition . The work of Stevenson ( 1 9 56 ) has
already been discussed , but this assertion of Birch ( 1 9 5 8 ) would explain
S tevenson ' s correlation of metabol ic activity increase on remoistening
dried soi l , with avai labil i ty of free amino acids and "other nitrogenous
materials " re leased by the air-drying process . S oulides and All ison
( 1 9 6 1 ) showed that when e i ther drying or freez ing was followed by a
period of incubation , there was an increase in the decomposition of soil
organic matter , this be ing substanti ally greater for drying than for
free zing . Pro longed drying increa sed the rate of decompos ition , and
mult iple dryings had a cumul at ive e f fect . Mult iple freezings had no
e f fect . I t would appear that the increased decomposition of organic
matter following intermittent drying or freezing is due primarily to the
re lease o f nutrients , e specially energy s ources that can be rapidly
oxidized by the soil f lora . From the above , much of these are likely to
be derived from microbes ki l led by the drying process . Providing the C/N
ratio is suf ficiently low to al low for the net ·mineralization of N , a
snowbal l effect result s , leading to further breakdown , and thus the burst
of C02 production and release of NH4 . following drying is enhanced by the
7 2
youthful state o f a growing ( feeding ) microbial population ( Soul ides &
Al l i son , 1 9 6 1 ) ; the microbes which survive the drying process are
provided with an excess of substrate, and so grow rapidly . The key to
succesful storage of soils therefore l ies in the maintenance of microbial
populations at f ield levels .
From the above i t is obvious that experiments - especially those such as
SNA measurements - should be carried out on soil in the f ield state . If
storage is necessary , then the period of storage should be as short as
poss ible and under condit ions of low temperature and high humidity .
Conditions of high humidity are easily attained when samples are stored
in sealed plastic bags at low temperature , especially when the soil in
the f ield is moist or even wet . Accordingly, all soil samples col lected
in this study were stored at 3 oc in sealed plastic bags . However , it was
considered desirable to check on the ef fects of a period of storage under
these conditions on SNA values , and gain some idea as to how long samples
could be stored .
Methods and Materials
This experiment was carried out in conjunction with the second experiment
of the f irst year of the work on the pH relations of nitrif ier activity
( Chapter 7 ) . Despite modif ication of the SNA technique to accommodate the
adjustment of incubation pH , the experimental procedure lent itself to
the study of storage effects s ince in addit ion to investigat ing the
ef fect of storage on the SNA value , it was also of interest to see i f the
relationship between nitrif ier activity and pH was affected by storage .
A bulk soil sample ( approx . 5 kg ) was dug from one of the control plots
from f ield No . 2 on Augus t 2 5 , 1 9 86 . The so i l was s ieved ( < 2 mm ) ,
thoroughly mixed, and a 200 g sub-sample leached overnight with . 1 dm3
0 . 00 5 M KCl a s bef ore . At the end of leaching , excess moisture was
removed from the soi l by suction filtration for 90 minutes after which ,
3 0 repl icate 5 g samples ( oven-dry equivalent ) · were placed into 5 0 cm3
incubation tubes containing 20 cm3 0 . 005 M KCl with 0 . 3 % �;v agar . The
pH of the suspensions was then adju:;ted in tripl icate by adding sma l l
7 3
amounts o f 0 . 1 M H C l or KOH . A pre l im inary experiment ( Chapter 7 )
established that the pH of the suspens ions attained an approx imately
steady value within f ive hours , and accordingly , after f ive hours , 1 0 cm3
0 . 0 1 M ( NH4 ) 2S04 was added to each tube and the incubations begun as
before . After the eight hour sampling, the incubation pH was measured by
glass electrode and pH meter .
Three weeks after this experiment , a further 200 g sub-sample of soi l was
taken from the original bulk sample which had been stored in a sealed
plastic bag at 3 oc and the experiment was repeated .
Results and Discuss ion
For both experiments , SNA values were plotted against pH and the data
were f i t ted w i th a quadrat ic equat ion by a l east- squares f i tting
procedure ( Figure 4 . 4 ) . By differentiating the . f itted equations , a pH
optimum for nitri f ication pHopt was calculated . The f i tted equations for
fresh soil ( Figure 4 . 4a ; R2 = 0 . 8 5 , p < 0 . 1 % ) and stored soil ( Figure 4 . 4b ;
R2 = 0 . 5 3 , p < 0 . 1 % ) were :
SNA = - 0 . 1 7 622 + 0 . 06864pH - 0 . 0 0 583pH2 ( 4 . 6 )
for fresh soi l , and
SNA = - 0 . 1 5 60 0 + 0 . 0 6 49 4pH - 0 . 0 0577pH2 ( 4 . 7 )
for stored soi l . Predicted values of pHopt were 5 . 89 and 5 . 63 for the
fresh and stored samples respectively . The cal culated SNA values at
pHopt 1 SNAopt were 0 . 02 6 v mol N03-N g- 1 h_ , for the fresh, and 0 . 02 7 v mol
N03-N g- 1 h_ , f or the stored soil sample . Looking at Figure 4 . 4 , there
appeared to be very l i ttle di f ference between the f itted curves for fresh
and stored soi l s . By combining analysi s of variance and analysis of
covariance ( Freund & Minton , 1 979 ) it was found that the SNA and pH data
for both experiments could be grouped and f i tted with a common quadratic
equation ( Figure 4 . 5 , R2 = 0 . 62 , p < 0 . 1 % ) , which predicted values of pHopt
and SNAopt of 5 . 76 and 0 . 0 2 6 respectively . i . e . the 3 week period of
storage had no ef fect on the SNA or the response of the nitrifiers to pH .
Accordingly it was concluded that samples could be stored for 3 weeks
without affecting experimental results .
a. Fresh soil
0.03
0.028
0.026
0.024
SNA 0.022 �mol
03-N
g-1 h-1 } 0.02
0.01 8
0 .01 6
0.01 4
0 .0 1 2
4 4.5 5
0 .03
0.028
0 .026
0 .024
0 .022
0 .02
0 .01 8
0 .01 6
0 .01 4
0 .01 2
5.5 6 6.5 7
pH
b. Stored soi l
*
*
* * * *
*
4 4.5 5 5.5 6 6 .5 7
pH
Figure 4 . 4 pH optima curves for nitrifier activity in ( a ) fresh soil and ( b ) soil that had been
s tored for 3 weeks
0.03 *
0.028 *
* "li(D {!) 00 0
0.026 * **
0
0.024
0
SNA 0.022 0 * (JJmol * N03-N
0.02 g-1 h-1 )
0
0.01 8 * CD
*
0.01 6
0.0 1 4 *
0.01 2
4 4 .5 5 . 5 .5 6 6 .5 7
pH
Figure 4 . 5 Co��on pH optimum curve fitted to SNA data for fresh and s tored soil
0 Fresh soi l
* Stored soil
7 6
A s mentioned above , Ross e t a l . ( 1 9 8 0 ) found no temperature to be
satisfactory for s torage of soil where maintenance of the biomass was a
priority , although 4 oc was adequate for short periods . In the l ight of
this and other f indings (detai led above ) , storage periods were clearly to
be avoided if possible . The period of storage for samples used in the
work reported in this thesis was never longer than 1 0 days , and was
generally conf ined to 3 days . Thus , on the assumption that any microbial ,
chemical or biochemi cal changes undergone by the soi l as a resul t of
sieving and storage were either negligible , or would have been manifested
over periods of s torage much longer than three weeks , it was concluded
that SNA me asurement s were una f fected by s torage , and therefore
represented f ield nitrif ier activity . Of course , there may have been an
immediate ef fect on nitri fer activity caused by sieving ; this was not
investigated , but wa s unavoidable since the soil . . had to be s ieved to
obtain a homogeneous bulk sample for the incubation experiments .
D . CONCLUSIONS
The prel iminary experimental work outl ined in this chapter was done with
the aim of tailoring the SNA technique to the Tokomaru s i l t loam , and
ensuring that the logist ics of the various parts of the technique were
compat ible w i th the intended l ines o f research . As a result , the
following was drawn up as a standard procedure for SNA measurements on
the Tokomaru s i l t loam , and was used as the basis of experimental
technique for the research reported in the fol lowing chapters of this
thesis :
1 . Following sampling, each soil sample was sieved ( < 2 mm ) and stored at
3 oc in a sealed plastic bag . The period of storage was never greater
than 1 0 days .
2 . Each sample was subsampled for moisture content and Ex-NH4. ( where
applicable ) , and leached with 0 . 00 5 M KCl in a Buchner funnel fitted
with a Whatman No . 1 fi lter paper at a soi l : solution ratio of 1 : 5 . ( In
the spatial variabil ity experiments ( Chapters 5 & 6 ) the leachate was
retained and analysed for N03-N )
3 . Excess moisture was removed by suction f i ltration for 90 minutes .
7 7
4 . 5 g samples ( oven-dry equivalent ) were placed into 50 cm3 plastic
tubes containing 20 cm3 0 . 00 5 M KCl with 0 . 3 % -!� agar .
5 . To each tube , 1 0 cm3 0 . 0 1 M ( NH 4 l 2S04 was added . The tubes were
sealed and placed in an enclosed end-over-end shaker f itted with a
thermostat , and incubated at 22 oc for 8 hours .
6 . Af ter 1 and 8 hours , a 5 cm3 sample of suspension was taken from each
tube , centrifuged at 3 0 0 0 r . p . m for 1 0 minutes and the supernatant
frozen and stored for N03 -N analysis at a later date .
7 . After the 8 hour sampling , the pH of the suspension was measured by
glass electrode and pH meter .
SECTION I I . AN ANALYSIS OF SPATIAL VARIABILITY IN NITRIFIER ACTIVITY
CHAPTER 5
VARIABILITY IN NITRIFIER ACTIVITY WITH DEPTH ��D D ISTANCE
A . DEPTH DEPENDENT VARIABILITY
7 8
It i s generally accepted that microbial activity i s higher in the upper
relative to the lower layers of soil profi les ( e . g . Speir et al . , 1 984 ,
Higashida & Takao , 1 98 5 ) . This is due to the f act that organic matter
( i . e . s ubs tra te ) enters the soi l system at or near the surface , and
consequently occurs predominantly in the surface layers , and the oxygen
needed by aerobic micro-organisms decreases in availabil i ty with depth
( Khyder & Cho , 1 983 , Colbourn et al . , 1 98 4 ) . When N is being mineralized
at a rate in excess of that of No3 - loss by leaching , one might expect
the distribution of N03 - down a prof ile to fol low that of NH4 + which ,
becaus e i t i s read i l y adsorbed by s o i l col loids , might i n turn be
expected to fol low the distribution of organic-N substrate . Cameron et
al . ( 1 97 8 , 1 9 79 ) found that both No3- and NH4+ tended to decrease with
depth in a clay loam in the range 0-60 cm , whi lst Young and Aldag ( 1 9 82 )
noted that only a very small proportion of the total soi l N occurred as
readily available mineral N . Stevenson ( 1 982a ) s tated that over 9 0 % of
the N in the surface layer of most soils was organically combined , much
of i t as amino acid-N or amino sugar-N ( Khan & Sowden , 1 97 1 , Stevenson ,
1 982b ) , al though the amount of amino-N as a proportion of the total N
tends to decrease with depth due to greater humif ication .
For the purposes of the prel iminary experimental work outlined in Chapter
4 , soi l samples were taken from the 3-9 cm depth range . As explained
previously , the top 3 cm were avoided to minimise any inhibitory effects
that grass roots may have on the rate of nitrification ( Molina & Rovira ,
1 9 64 , Neal , 1 969 , Moore & Waid , 1 97 1 ) . However , the assumption that this
would be the most suitable depth range of sampling for an invest igat ion
of spatial variability in nitri f ier activity in the Tokomaru s i l t loam
79
may not have been a good one becaus e ( a ) grass roots may not a f fect
nitr i f ication in this soi l ; ( b ) nitrification may occur at higher ( or
lower ) rates in the 0 - 3 cm range regardless of the effect of grass roots ;
and ( c ) the degree of any spatial dependence in �itrif ier activity may
not be the same at a l l depths . As a consequence , a f ield-scale estimate
of the mean so i l n i trate concent rat ion , made for use as an input
parameter to a nitrate leaching model may not be a bes t es tima te due to
depth dependence . Thus , i t was considered important to investigate the
vertica l distribution of nitrif ier activity and i ts associated parameters
in addition to any spatial analysis .
i . Methods and Materials
Soil sampling
Ten sampl ing sites were selected in f ield No . 6 us ing random numbers to
generate the coordinates o f each s i te . Fol lowing adjustment ( where
necessary ) of the posit ion of some of the si tes to ensure a minimum site
separation of 3 m , the soi l at each site was sampled to a depth of 24 cm
in 3 cm layers using the corer described in Chapter 4 . Two immediately
adjacent cores were taken at each site . The soi l from the two cores was
sieved and bulked on a depth basis to give 80 samples representing the 1 0
sites at 8 depths . The samples were stored as described previously prior
to analysis for SNA and exchangeable-NH4 . The leachate from the SNA pre
leaching was retained for analysis of N03-N to give the quantity of N03 -N
present initially per g soil .
A further f ive sampl ing s i tes were randomly sampled in the manner
described above ( only one core per site ) and the soil sieved and stored
as before . These samples were analysed for their total nitrogen , carbon
and phosphorus contents following the methods described below .
S i x addit ional sites were randomly selected for the measurement of soi l
bul k dens i t y . ( A m in i mum of f our samples a re requi red f or thi s
measur emen t ( D . R . S co t ter , Dept . S o i l Science , Massey University -
personal communication ) , but six si tes gave an easily manageable number
8 0
and a l lowed for more precise estimation o f mean values - see Chapter 3 . ) .
At each of these sites , cores measuring 5 cm deep with a diameter of 4 . 8
cm ( i . e . 90 . 4 8 cm 3 soi l ) were taken to a depth of 25 cm , oven-dried
overnight at 1 05 oc , and the bulk density calculated on a g dry soi l cm- 3
basis .
Sampling was carried out during mid-May ; there was a period of 9 days
between the sampling for SNA measurements and that for ana lysis of C , N
and P . I t was assumed that any change in the latter soil properties
between the two sampling dates was insignif icant .
Analysi s
SNA , exchangeable-NH4 and N03 -N were analysed by the methods described in
Chapter 4 .
The ana lysis of total nitrogen and phosphorus ( 4 replicates per sample )
was carried out by Kj eldahl digestion fol lowing the method o f Bolan &
Hedley ( 1 987 ) . 1 g f inely- ground air-dry soil was placed in a pyrex tube
and 4 cm3 digest acid ( 25 0 g K2S04 and 2 . 5 g Se powder dissolved in 2 . 5
dm3 cone . H2S04 ) were added , and the tube heated at 350 oc for 4 hours .
After cooling , the contents of each tube were di luted to 5 0 cm3 wi th
deionised water , thoroughly mixed in a vortex shaker , and the solutions
s imu ltaneous ly analysed by autoanalyser for the ir N and P contents ,
following the method of Twine and Will iams ( 1 9 7 1 ) .
The carbon content of each sample was analysed by dry combustion in a
stream of 02 using a Leco furnace fol lowing the method of Bol an and
Hedley ( 1 987 ) ; a copper oxide catalyst was used to promote the conversion
of CO to C02 , and an Mn02 trap used to remove any halogens present . The
amount of carbon in the soi l was calculated as the mass of C02 produced
( mg ) x 0 . 2729 ( the proportion by mass of C in C02 ) per mg air-dry soil .
8 1
i i . Results
Figure 5 . 1 a shows the bulk density at each s i te plotted as a function of
the depth at the centre of each core ( the centre of the 0 -5 cm depth core
for example , was taken to be at 2 . 5 cm depth ) . Since the dimensions of
the bulk density corer were dif ferent from those of the corer used for
all other soil sampling , it was necessary to f ind an expression relating
bulk density to soil depth so that the bulk density at the depth at which
the other samples were taken could be interpolated . The data followed a
curvil inear trend with depth and were best f itted using least squares
optimization by the equation ( R2 = 0 . 88 ; p < 0 . 1 % ) :
0 . 9 3 6 1 1 + 0 . 0 355D - 0 . 0 0 0 64D2 ( 5 . 1 )
where p b is the bulk density and D denotes depth ( cm ) . It should be noted
that this equation is valid only over the depth range 0- 2 5 cm .
The soil moisture content at the time of sampl ing for SNA analysis was
calculated from the moisture content of the sieved soil us ing equation
( 4 . 5 ) , and these gravimetric data were converted to volumetric moisture
contents , S v , using values of p b calculated from equation ( 5 . 1 ) . The data
at each depth were assumed to be normally distributed ( see section B , and
also Chapter 6 ) and the mean values are plotted against depth in Figure
5 . 1 b .
The spatial analysis ( see section B and also Chapter 6 ) indicated that
SNA , in it ial N03- and Ex- NH4- conformed to log -norma 1 distributions .
White e t a l . ( 1 9 87 ) suggested that for lognormally distributed data , when
the variance of the natural logarithms of the property values i s less
than 0 . 5 , and the number of samples is large , I< the best estimate , � ' of
the mean , � ' of the population from which the sample is drawn is given
by :
"
� = x .. exp (� + l) / 2 ) ( 5 . 2 )
a. Bulk density b. Moisture content
g cm-3 cm3 cm-3
0 .9 1 1 . 1 1 .2 1 .3 1 .4 1 .5 1 .6 0 .2 0 .3 0 .4 0 .5 0 .6
0 0
3
5
6
1 0 9
Depth 1 2 (cm)
1 5 1 5
1 8
20
2 1
25 24
Figure 5 . 1 Change in ( a ) bulk density and ( b ) mean volumetric moisture content with depth in the Tokomaru
silt loam sampled in mid May
8 3
where Xa i s the estimate of the mean value o f the sampled property , and �
and V are the arithmetic mean and variance of the natural logarithms of
the property . Both SNA and soil No 3 - f u l f i l l e d the criteria for use of
equation ( 5 . 2 ) with respect to V and despite there being only 1 0 samples
for each depth , this equation was used to est imate the mean values . In
contrast , the va lue of l) in the cas e of Ex-NH4 ... exceeded 0 . 5 at all
depths and so the mean value was estimated using Sichel ' s est imator , XQ
( Sichel , 1 9 5 2 ) where :
x.. = { exp ( � ) } { 1 + l) + ( n - 1 ) V 2
2 2 22 ! ( n + 1 )
+ ( n 1 ) 2 V 3
233 ! ( n + 1 ) ( n + 3 )
+ . • • } ( 5 . 3 )
The data f or C , N , P and C / N were as sumed t o con form t o normal
distributions as did the incubation pH ( see section B and also Chapter
6 ) . The distributions of all the measured properties and their standard
error a t each depth are shown in F igure 5 . 2a -h . In the case of the
normally distributed properties , the standard error , S . E , was calculated
using the equation ( Clarke , 1 98 0 ) :
( 5 . 4 )
where S2 i s the sample variance , and n is the number of samples ; the
vari ance of the s ampl e mean i s g iven by s2 /n . Whi te e t al . ( 1 9 87 )
presented equations for the variance of both Xa and XQ and used these to
infer the rel iability of Xs relative to Xa . However , it i s suggested that
for the purpose of est imating standard errors of lognormally distributed
properties , dist inguishing between the variance of Xe and Xs may not be
necessary . S ichel ( 1 95 2 ) stated that the variance of the estimate of the
arithmetic mean Xa , denoted here by Var ( X ) , could be calculated using the
equation :
Var ( X ) = { exp ( 2� + V ) } { [ exp ( V ) ] [ 1 - 2V ] - < n.- , > / 2 - [ 1 - V ] - < r> - ' > } ( 5 . 5 )
n n n n
a. SNA b. In itial N03
0
3
6
9
Depth 1 2 {cm)
1 5
1 8
2 1
24
0
Figure· 5 . 2
pmol N g-1 h-1 pmol N g-1
0.01 0.02 0.03 0.04 0 0.2 0 .4 0.6
0
�
3 / I /
_.. 6 oil
... 6 I ... ...
J. qY .....
9 ..... I ..... .....
@ .....
.....
1 2 ..... .....
.....
$ .,... I
I 1 5 I
9 rf
/ 1 8 \ /
� �
\ 21 I
� L-..j.....
24
Depth profi les of ( a ) SNA , ( b ) N03 - , ( c ) Ex-NH4
+ , ( d ) incubation pH , ( e ) total carbon ,
( f ) total nitrogen , ( g ) C /N ratio , ( h ) total phosphorus and ( i ) % mineral N i n the
Tokomaru s i lt loam sampled in mid May
c. Ex-NH4
------ --- -- --
pmol N g-1
0
3
6
9
Depth 1 2 (cm)
1 5
1 8
2 1
24
0 0 .5
/
'
'
'
/
/
....... ....... ..... .....
Figure 5 . 2 ( Contd )
1 1 .5
d. Incubat ion pH
pH
5 5.2 5.4 5.6
0
-"""