1 Introduction Precision agriculture is based on variable managem-ent composed of describing and understandingvariability,variable- rate technology and decisionsupport systems (Gupta et al,1997;Shibusawa,1998;Robert,1999;L i et al,2 0 0 0 ) . Describingtemporal and spatial variability of the field parametersis the first step for precision agriculture.Those fieldparameters involve soil parameters,yields,weeds,insects and so on. Grid sampling was the most common way for soilsensing.It divides the test field as a number of gridswith the same size,and then only collect one soilsample from each grid to analyze soil properties overwhole field (Burgess and Webster,1980 ;Burgess andWebster,1984 ;Webster and Burgess,1980 ;Websterand Burgess,1984 ) . In these cases,it was a tacitagreement that one sample collected from a grid couldrepresent the whole characteristic of that grid.However it is obvious that estimation error isunavoidable.Sometimes it might be very big if onlycollecting one sample from a grid. Thus it becomesnecessary to collect more samples from a grid in orderto decrease estimation error. Here the number ofsamples collected from the same grid was calledsample size (Kvanli,1988) .According to statistics ifsample size is big,the estimation error will be small.However,it is impractical to collect too many samplesfrom a grid because of budget constraints(both timeand money) .We should make a compromise betweensample size and estimation error. Hence,it isnecessary to understand whaterror would be producedwhen only collecting one sample from a grid.And itisalso necessary to determine how many samples shouldbe collected for representing the general feature overone grid space(Khan and Nortcliff,1982 ;Pardo etal1998;Sakai et al,1992 ;Tsegaye and Hill,1998) . The content of nitrate nitrogen (NO3 - N) ,as a soilparameter,is very important since it is not only afertility index,butalso a pollution resource to soil andunderground water.Hence many studies have beenconducted to confirm its variability in a field(Jaszberenyi et al,1999;L i et al,1998;Li et al,1999;Shibusawa,1999) . After grid sampling,variability of NO3 - N was analyzed and soil mappingwas performed. The aim of this study was to clarify the relationshipbetween sample size and estimation error.That is,toinvestigate what error would be produced when onlcollecting one sample from a grid,and to calculatehow many samples should be needed in order to keepestimation error small.2 Materials and methods The test field was a growing cornfield on theuniversity experimental farm,Tokyo University ofAgriculture and Technology. Soil texture wascomposed of2 0 %of clay,34%of silt,and4 6 %ofsand,which was classified into CL .Test area locatedin the central part of the cornfield,was 4 .2 m× 4 .2m.It was supposed to be a grid of grid sampling. The test area was further divided into 4 9sub- gridswith the size of 0 .6 m× 0 .6 m.And from each sub-grid one sample was collected in each test.Thesamples obtained from all sub- grids were supposedfrom the same population,and were used to thisstudy. Seeding was performed on May 12 ,and the firstsoilsampling was carried out a month late after seeding.The test schedule was shown in Table1.For topdressing,the contents of N,P,and K were 6 .0 %,3.0 %,and 6 .0 %,respectively.The quantity was 32 4kg/ hm2 .Table1 Sampling schedule of the testDate May 1 2 June 1 2 June 1 8June 2 4 July 0 4 July 1 4 July 2 4 August0 4 August 0 5 August 1 4Farm work Seeding sampling Top dressing Sampling Sampling Sampling Sampling Sampling Harvesting Sampling Single core was collected within a sub- grid in eachtest.The core was1cm in diameter and 30 cm indepth.In order to study the difference of sample sizeamong different depths,the top 0～ 10 cm,the middle10～ 2 0 cm,and the bottom2 0～ 30 cm of each corewere collected respectively for analyzing NO3 - N,i.e.3× 4 9samples were collected in each test,and finally10 2 9samples were collected in all seven tests.NO3 - Nwas analy
More abstracts about the 土壤硝态氮含量检测和分布图生成中的最优样本量(英文)