:‘ILITJ‘ 1‘1.'I-4 ‘-I . I ‘ .1 . , A ‘11 , I”. ,_ ILA-II 1) III?” 1,1,, II ~‘,,,"I~ . . , .4 .., ',,,.II - ' I ,3“ II. " 1 1" '1‘. «.41 .1qu ' . 'MJ‘I :I‘,,‘,-W¢ 0 11‘“ .3121."- I‘m ,..411 v1.11, ,4 I..,,'I‘,‘J“4l' . . .H ‘ “.4 f- .. I‘I'.I.‘ I‘ ' ”2-?th "4,“, ,.I11 . I . I‘,: ,. 9;,“ ‘1‘ “If“ :“1‘II'1‘Mr, ‘. , ,II‘I!‘ EH, 1 ,4; I‘“ g‘IF MUM” "WI 1 I31..1.I ,1,‘ II1I V. ‘sz [I] I .I ’ 5. .0111 .IIII 1, I IIIIIL.“ “11,11," I‘I .1." .‘ .11.~- , , 4 ‘\ IKWHXI 0.1/1" jde‘Ih 7IIII1tIII'IIII'II.IIIIIIIINVNQI 9*; III IIIIIIIUII III-III: Ian *’1 In“: 1,1,1 . . . I A}: {:4} , ‘.1 I, . ‘1. ' . , ’ . . '1.- I, ' , . ‘ , "J'I'I‘IIII 2.1.11.1, 1... 1., M 1,1 .IQEA‘I .WI-1I1,' ' 11.,1_1“ 1,3“, ,. ,I1}, "/15. 1,44,, -1,.1I n- “q: 1r1dvfl-1J"‘ , - 115§h.v%1411,11,l'fi:|\. . .' . :"Y' 1;!1E, 1. 1- '3‘: i1 IIIII'lfO'III .111}; .-,:.,'.,1‘ '11 >I'I‘ II._I,, "' I-I‘I'II , l' III'IIg‘; ' [1.1” ,,,.1‘: , ,, -. 1., a: J“ .' ‘1 1 1 : .11”? .. 3.113%, 311,112.. 4,3. ufl“ ' 111,111,131) 5., 1:. IIIIIIII‘ I' 'IIIfit’Ifi111Lah1whbfmfi. 1:11;;1-1111‘ 1511‘?- ,5; 1.1,, 1:. 1‘12?" .“"1I‘ I‘ ' 1‘ ,3'1'1‘3" 2,13,,111,” h .1:ng . ' I I, I 9,111 - ' L», ,. I511" :41“ :1,” , 121‘ I1“ . IT1.:,I'." 1‘11 “ ‘I'I ,l 11‘“ ,ItI. ,. l- .‘I I ,3‘1.11‘..1',1.J,—, ,1, 7‘ III '1' th I. I12) ”I 1' "' .'1'.,,"',..U.',I.I1I1_ ,jI fifl‘fi‘f .‘ HI. (1'11. WWI' .11. ‘ I ,'.|'-'- 511,111.}. 13?”... . 12‘“ -.1I:II‘ “ .-. I."I' “ .1.I1I15‘II‘}',I .2 I J_ l 1 g E. g . ' . ‘3 ~ W Zmrrm if“ . 257'400 257'450 257'500 257550 25T600 257350 'V‘l Easting (m) Figure 1.1. The Order two soil survey map (1:20000 scale: Threlkeld and Feenstra.(1974) (A) and a high resolution Digital Elevation Model (m) for the experimental field (B). A 4747 A 4747 g. _. '- , , - .. E a a»? ' . U) 5 ~ . .C l .t g Bi: new? r y’- 4: .;. ~ 47476 ' fl-é‘?’ it" 47475 " "m " - 257200 257300 257400 257500 257600 257700 Easting (m) jg>>z 4747900~ 4747850— 4747800-I '7 257400 257450 257500 257550 2576001257650 Easting (m) Figure 1.1. The Order two soil survey map (1 :20000 scale: Threlkeld and Feenstra,(1974) (A) and a high resolution Digital Elevation Model (m) for the experimental field (B). 12 Table 1.1 Map unit descriptions, area within the experimental site, and population of grid samples falling within map unit for the experimental field in Durand, Michigan. Soil Mapping Unit Map Map Soil Landforms Parent Material # of Unit Area Grid Symbol (%) Points Breckenridge Bt 19 poorly drained, sandy loam material, 8 sandy loam nearly level soils 45 to 110 cm thick, 0 to 2 % slopes on till plains overlying clay loam glacial till Conover loam, CtA 15 Somewhat loamy glacial till 7 0 to 2 % slopes poorly drained, nearly level to gently sloping soil on till plains and low moraines Macomb loam, MaA 12 somewhat poorly loamy material 6 0 to 2 % slopes drained, nearly level or gently sloping soils on till plains and low moraines Metamora sandy MsB 19 somewhat poorly sandy loam, 45 to 100 12 loam, 2 to 6 % drained, nearly cm thick, and in the slopes level to gently underlying loamy sloping soils on glacial till till plains and low moraines Wasepi sandy loam, WeA 35 somewhat poorly gravelly light sandy 20 0 to 2 % slopes drained, nearly clay loam and sandy level to gently loam, 60 to 106 cm sloping soils on thick, over gravelly outwash plains, coarse sand lake plains, stream terraces, and glacial drainage-ways 13 The landforms descriptions (Threlkeld and Feenstra, 1974) for each map unit are given in Table 1.1 along with other information about the composition of that map unit within the experimental site. A 30.5 m regular grid was imposed on the experimental site (122 by 305 m, 3.72 ha) in May 1995 (Figure 1.2a). Intact soil profile cores (5 cm i.d.) were obtained from each grid point to a depth of 1 m away from the wheel tracks using a hydraulic probe. When compression on the soil profile samples was suspected, the entire sample was discarded. Cores were placed in PVC pipes that were split to accommodate the core, reassembled, taped and capped to protect them during storage at 4° C. Each soil profile was sectioned by soil horizon and its thickness recorded. The top 5 cm of the Ap horizon of each soil core was removed and ground, and total organic carbon was determined by dry combustion using a CHN analyzer (Carlo Erba Instruments, Italy). Small sections (5- 7 cm length) were cut from each horizon within each soil profile, coated with plastic, and subjected to standard procedures for determination of soil bulk density (Blake and Hartge, 1986) and water retention at -33 kPa (Klute, 1986). The remainder of each horizon was carefully ground and the soil analyzed for particle size analysis by the standard hydrometer method (Gee and Bauder, 1986) and water retention at -1500 kPa (Klute, 1986). Water table depth was measured during spring and summer of 1995 using piezometers installed at each grid point to a depth of 1.5 m. Intact soil cores (7.6 by 7.6 cm ) were obtained in triplicate from 0-7.6 and 7.6- 15.2 cm depths at each grid point using a double-cylinder, hammer-driven core sampler (Blake and Hartge, 1986). .4. . Au- .. . . . A . N 4747...‘ O O O O O O O O O A CtA .2 24 25 26 27 28 9 30 31 32 4 A O O O M B O O O O 0 O 5 4747 s weA g 1 13 14 15 16 17 18 19 20 21 .r E 0 LA 0 o o o o o E Bt z 2 3 4 5 . 7 8 9 10 11 4747." 0 - . . . . . . . MaA ., 46 47 ; 49 50 51 52 53 54 a O O O-) O O O O O 0 257400 257450 257500 257550 257600 257650 Easting (m) 1 [’1\ A l 1 B WeA E CIA M53 03:: afogfl 3 47478504 I: 8 tu- 8" :8 _ "E 00‘... O 00 o 0 2° . , . lo 0 0 Bi .0 4747600I u- 3 t I '0 o . #400 #450 257% 257% New 257550 Easting (m) [- 1 A leB “A m: : :3: a A : : a; Q o . 1; 4747850 5. r ; i E 5" 0 . "E II o .‘9 : 2 2 g 3 I V o . g 9.: t 9 474760011 ° g I'- '1 0 6 . 257400 257450 257500 257550 257600 257650 Easting (m) Figure 1.2. Location of the 30.5 m regular (A), geospatial analysis (N=162) (B), and geospatial analysis (N=419) (C) gn'ds relative to the map units from the Order two soil survey. Refer to Table 1.1 for map unit descriptions. 15 The cores were taken away from the wheel track and 20 cm from the corn row. Following collection, cores were stored at 4° C until analysis of soil physical properties. The soil cores were saturated from the bottom for 48 h and soil water retention was determined in pressure chambers at -6, -33, and -100 kPa matric potentials (Klute, 1986). Total porosity was calculated using the water content difference between saturation and oven-dry states. Cores were oven dried for 48 h at 105° C and bulk density was determined as the mass of dry soil per volume of field-moist soil (Blake and Hartge, 1986). Bulk densities of individual cores were used to calculate soil water retention as volumetric water content. Air filled porosity was calculated at each matric potential measurement. Data from the three cores were averaged to provide a mean value for each sampling depth at each grid point. Two other sampling events were established to compare the effect of sampling design on the spatial analysis of soil physical properties. A single, intact soil core (7.6 by 7.6 cm; N=162) was obtained on 13 to 20 July 1995 from the surface 7.6 cm within a 100 by 100 m sub-area of the experimental site as illustrated in Figure 1.2b with sample locations spaced at distances ranging from 1 to 100 m. Cores were taken away from the wheel track and 20 cm from the corn row. Water retention, total porosity, air-filled porosity and bulk density were determined as above. Bulk soil samples (N=419) were obtained on 10 July 1995 from the Ap horizon within a 100 by 300 m sub-area as illustrated in Figure 1.2c with sample locations spaced at distances ranging from 1 to 300 m. Particle size was determined on each sample as described above. Table 1.2 lists all variables with the respective symbols used in this study. 16 Table 1.2. Description of variables with the respective symbols used in the study. Symbol Description IGF Indicative Goodness of Fit. IGF is a number without units; a value close to zero indicates a good fit. C0 nugget efi‘ect C, structural variance R Range (m) S Spherical G Gaussian E Exponential L Linear PN Pure nugget a a pure nugget effect was found and therefore no sill or range could be determined. CV Coefficient of variation (%) SD Standard deviation Bt Depth to the Bt horizon (cm); Ap Thickness of the Ap horizon (cm); Ap—1500 Gravimetric moisture content at -1500 kPa matric potential (%) of the Ap horizon TC Total carbon content in the top 5 cm (g kg") P-clay Weighted average profile clay content (%) P-clay+silt Weighted average profile clay + silt content (%) PAW Weighted average profile available water (%) AWG-Ap Gravimetric available water content in the Ap horizon (%) AWV-Ap Volumetric available water content in the Ap horizon (%) GM Gravimetric moisture content (%) VM Volumetric moisture content (%) AF P Air filled porosity (%) BD Bulk density (Mg m") TP Total porosity (%) -6, -33, -100 kPa matric potential. WT-MIN Minimum depth (m) of the water table. WT-MAX Maximum depth (m) of the water table. 1 7 Statistical Analysis Geostatistical analyses were performed using Variowin 2.2 sofiware (Pannatier, 1996). Variograrns were derived by plotting the semivariance of the soil data as functions of vector distance. Each parameter was subjected to semivariance analysis using isotropic and anisotropic models (linear, spherical, exponential, and gaussian) for defining semivariograms. The models were then used along with kriging techniques to develop maps showing spatial patterns in variability of selected soil properties (Issaks and Srivastava, 1989). Descriptive statistics and correlation were calculated using SAS (SAS, 1990). RESULTS AND DISCUSSION Map Unit Purity The first phase of this analysis was to examine the variation within map units by comparing soil profiles obtained on a 30.5 m grid with map unit descriptions in the published soil survey. The map in Figure 1.3 provides the location of the grid soil profile samples relative to the map units from the Order 2 soil survey, with each profile numbered for reference. For each map unit, Figures 1.4-1.8 compare the horizons and soil textures of the representative soil profile as given in the soil survey of Shiawassee County (Threlkeld and Feenstra, 1974) with the grid profiles falling within the map unit in the experimental area. The characteristics of the Ap horizon, the Bt horizon, and the C horizon of each grid profile were compared to the corresponding horizons of the representative profile of that series. 257200 257300 257400 257500 257600 257700 N Easting (m) Figure 1.3. Location of the grid soil profile samples relative to the map units from the Order two soil survey. Refer to Table 1.1 for map units descriptions. l9 9 1011 20 21 22 53 54 0cm — '_—‘ '_ —' —‘ —l Ap APQJAPAPAFAPQ sl 25_ 3' Bt siJ 1 1' Eg 5' h; i sl 1 Es: _ ‘E _ __ _ B 2B1 B cl AB sl sl (1 E sl scl Bt __ __ Is 50 BE 2312 — sl 3‘ sl Bt] sl ._ s] sci _ 33d at"id/11 ——. inn sl ' 5 ml Bt sl Btg — Bti. I 75— 8d :- Sclll sl sl 2C 1cl/s Rt 2_ L— 3; g . g BC BC 4 sl/scl __ ls _ BC S] t [fin 1' _ 1' C3 iBC ”35 C d J ls l w I 1 I l 100.. ls __l ‘EB‘J’B i i E Figure 1.4. Representative soil profile of the Breckenridge sandy loam 0 to 2% (Bt) as given in the Soil Survey of Shiawassee County, MI (Threlkeld and Feenstra, 1974) and the grid profiles falling within the map unit. Refer to Table 1.1 for map units descriptions. 20 1 12 13 23 24 34 35 36 37 38 00m _ _ _ _ _ _ _ —1 —- -— r— r— M MMMMMMMMMM 1 s1 81 254 E —— 111.3591 sl rs]— g BtlsEBE “El—El —l E E — l E »— 91 Al Btl sl -— Btg _‘ t 31 8' sl F sl — sl/l — m— — nm‘p 50 .1 _ 3.21... “$11.81; 3.1 -— Bt BC _ Btll L s] sl sl 1C1 ' 91 5c — sl "_‘ — _ A: _ BC Btlsc 3‘ l m 1 C1 __ 75_ Is _ r— l scl ls C 2C C L_ _ Bel letZ sl 2 )— BC cl C BtJ il/ls ch siE‘Isiscisifljssc 100 _' .i .51 _i .4 J J J J Figure 1.5. Representative soil profile of the Conover loam 0 to 2% (CtA) as given in the Soil Survey of Shiawassee County, MI (Threlkeld and Feenstra, 1974) and the grid profiles falling within the map unit. Refer to Table 1.1 for map units descriptions. 21 14 15 l6 17 25 26 27 28 39 0cm __ fl _ j _ _ _ _ .1 Ap All Ap AHA An Ap Ap ApA ls s] l i] :11 fl; 1 l 25_ E 1E1 sl 1311,] g 1' g a 1E 180 my sl [Bl‘i sl 31 E — 5 BE sl Es 11310 E2 i— lBt scl ,—1 S] at — 50— sl 1' 1 fi 1 "mint — S c t — C1 — EB B Is 123 8| — t —— C1 1 SC Bt] sl — cl _ BC sl — IBC sc BC 75 ZBtg — H— C scl 51 :23 C sl s “1 s 1 cl cl _ —- i— C2 _ lBt ht C2 C ZCg ”L . iLflflflflflflfl Figure 1.6. Representative soil profile of the Metamora loam 2 to 6% (MsB) as given in the Soil Survey of Shiawassee County, MI (Threlkeld and Feenstra, 1974) and the grid profiles falling within the map unit. Refer to Table 1.1 for map units descriptions. 092 25_ 50_ 75_ AP 31 Bg Bt scl 2Btg cl 100_ 2C 22 2 3 44546 47 —. An El sl E2 Bt J_ *1 A1) hula 0 | AP Cl '—1 An :zci. fl] —. B — Bt: —1 A1! Ft] ,— .— 2C1! __ Arr ht} Bti. sl l: sl BC j_ Figure 1.7. Representative soil profile of the Macomb loam 0 to 2% (MaA) as given in the Soil Survey of Shiawassee County, MI (Threlkeld and Feenstra, 1974) and the grid profiles falling within the map unit. Refer to Table 1.1 for map units descriptions. 23 0cm 567 81819293031323340414243444849505152 —— AP AJI; A A A A A AJlA—“A A.- A A A A A A A A SI SI SI SI 25 Eg L s E t sl E sl Bt l i0 g ls c] c C t scl .15 3‘ sl 2C 1 lllllllslslll3 s sc s c s s s s 100 it” i is i I 7 7 i ls Figure 1.8. Representative soil profile of the Wasepi sandy loam 0 to 2% (WeA) as given in the Soil Survey of Shiawassee County, MI (Threlkeld and Feenstra, 1974) and the grid profiles falling within the map unit. Refer to Table 1.1 for map units descriptions. 24 The Ap horizon of the entire experimental area was characterized as a sandy learn texture regardless of map unit, with a few profiles having textures of loam or loamy sand (Figure 1.9). The Ap horizon was generally consistent with the map unit descriptions for the Breckenridge sandy loam, the Metamora sandy loam, and the Wasepi sandy loam. The series description for the Macomb loam indicates a sandy loam texture but the map unit description for Shiawassee County indicates the surface to be loarn texture. The surface of the Conover loam should be a loam texture but this was not the case for any profile within the map unit (Figure 1.5). The clay content of the Ap horizon is within but at the low end of the range for a loam texture (8 to 27%), but there is too little silt and too much sand (maximum sand for a loam is 52%) to be classified as a loam texture (Figure 1.5). However, the map unit description for the Conover loam does indicate that within this survey area, the map unit includes some small areas where the surface layer and subsoil are sand or loamy sand. The representative profile for each soil series indicates the presence of a Bt horizon and that some portion of the Bt is gleyed. Few evidences of mottling were found in the profiles analyzed, although a water table was present within the soil profile at many of the grid locations during spring and summer of 1995. The texture, thickness, and depth to the Bt horizons varied greatly among grid points within each series and differed from the representative profiles of all series (Figures 1.4-1 .8 and 1.10). Areas defined by the soil survey as till plains (soils formed in loamy glacial till materials) contained typical outwash parent material, i.e., profiles 35 and 13 (Conover loam), 16, 25, and 26 (Metamora sandy loam), 45 and 46 (Macomb loam) (Figures 1.5, 1.6 and 1.7). 25 A N mmwvvavefi “"‘v A stagggazizgzsgzo‘e \W ’5 '5: \W/ / . 4” n 4747850 . . //// E , . f ..J°.3°3°3°3°3°.,,;. ‘15 “’r~443:3?zztzzztztttzo’ao's ~ 0 so oo.¢§‘9,¢.o.qp¢:§ ‘00 z ”3:: :wvge 03%, ,t 474m— :QQ..’: ._ v. . .:::§::.‘ ' ‘~,:>’;o.o:o:o:o:."::.‘\,o.o, ot’ . 3:44.83 43:48:. / 257400 257450 257500 257650 Easting (m) 1 I 1 \ A1 ‘ 1 1 B E 3; 4747850 _ .s ‘15 0 2 4747800 257400 257450 257500 257550 257600 257650 Easting (m) 4747900 I A A1 1 1 C E ‘5 4747850- — .5 € 0 z , 4747800~ — 257400 257450 257500 257550 257600 257650 Easting (m) Figure 1.9. Interpolated maps of clay (A). silt (B), and sand (C) contents (%) of the Ap horizon for the 30.5 m grid. Refer to Table 1.1 for map units descriptions. 24 26 M w 4 9 4 A 0 O 5 5 6 6 W :75 2 2 m m 6 6 .v. v. . 2 2 0 w 5 15 in ‘70 ) R20 ) 2 m m A ( g g I n .n 000406.9\ 0 was 0 w . 090%. 0 0 . $0940., 00. 5 a 5 a 20.5 I. .7 F. v. 7 E 5 .. . . 5 1 y 2 2 0 E E v . A 4‘ . . ,. _ : ... .‘éeoe‘\ a 0 o .1‘ /‘Q‘ 4040‘ 00 ‘ ‘ . 0w _ .4é4s4ysm 7 .0». 0% $ .0109 § \\ \ 5 $4595.... 5 a7\\\\\‘ 2 o $ . ;%5. 2 w. , w, m m m m m m m m m m m m 4 4 4 4 4 4 :5 82.52 NA as 9:562 EC 9:562 Easting (m) Figure 1.10. Interpolated maps of clay content (%) of the Bt horizon (A), thickness of the Bt horizon (cm) (B), and depth to the Bt horizon (cm) (C) for the 30.5 m regular grid. Refer to Table 1.1 for map unit descriptions. 474 474 Northing (m) 4747 27 Northing (m) 47478004 Northing (m) 257400 257450 257500 257550 257000 257650 Easting (m) V 0 0 0 0 0:0:0’4 0.0.0.04 0‘ 0 0’0’0‘ . 1.0.0.0 O 0’ k” 0 ' 0 0 ’0 0 030.0.0 0 v V 9 0:0’3’0’0 ’0 O b 0 0 0:0:0z0z0 Qfiék&fi§h KQXQKQMVIIIA gs p‘. 0 W5 06. is“... 257500 ‘ V - N m . ‘1 .5 (II 0 257550 Easting (m) jZZ§\A~Aa§§?3 , / 8'. ‘ '0' \V» r. / » g... 257600 257650 [15 Easting (m) Figure 1.11. Interpolated maps of clay (A), silt (B), and sand (C) content (%) of the C horizon for the 30.5 m grid. Refer to Table 1.1 for map units descriptions. 28 In addition, typical till materials were present within areas described as outwash parent material (soils formed in gravelly light sandy clay learn and sandy loam, 60 to 106 cm thick, over fine gravelly coarse sand) by the soil survey, i.e., profiles 31, 32, 41, and 29 (Wasepi sandy loam) (Figure 1.8). The texture of the C horizon for the grid profiles did not generally match that of the representative soil profiles (Figure 1.11). The C horizon of the Bt map unit was supposed to be clay loam but the profiles had a mix of sandy loam, loam sand, and sandy clay loam textures (Figure 1.4). Only 2 of 10 profiles (20%) within the CtA map unit (Figure 1.5), only 1 of 9 profiles (11%) within the MsB map unit (Figure 1.6), and only 1 of 6 profiles (17%) of the MaA map unit (Figure 1.7) had the expected loam texture in the C horizon. The C horizon of the profiles within the WeA map unit (Figure 1.8) had a mix of sandy loam, loamy sand, loam, sandy clay loam, and clay loam textures, not the gravelly coarse sand expected from the series description. Where the C horizon was supposed to be a loam, there was too much sand and where it was supposed a sand the soil had more clay and silt (Figure 1.11). None of the grid soil profiles appeared to match the map unit description for the soil series within the experimental area. Either there are many inclusions within this area and these were picked up by the nature of the grid sampling design or the soil map units do not correspond to the soils within this area of the field. These data suggest that the current soil survey map will be of little use in delineating soil management zones in this experimental area. 29 Spatial Analysis of Soil Physical Properties The second approach to establishing soil management zones for SSM was to analyze the spatial dependence of soil physical properties within this experimental area. The interest here was whether the soil profile and surface core data from the 30.5 m regular grid would provide accurate maps of soil physical properties that could delineate useful soil management zones within the experimental area. Also how sampling soil more intensively with more complex sampling designs improved the estimation of the spatial dependence of selected soil physical properties was tested. Spatial Analysis for the Regular Grid The data for each 30.5 m grid soil profile were parameterized to provide new quantitative variables for analysis as given in Table 1.3. These profile variables and the soil properties measured on the intact soil cores obtained from each grid point were subjected to geostatistical analysis (Tables 1.3 and 1.4). Maps for these variables interpolated using the calculated semivariograms are given in Figures 1.12-1 .19 with overlays of the soil survey map units provided for reference. 30 Table 1.3. Descriptive statistics and parameters for variogram models of variables obtained from undisturbed soil profile cores (5 cm i.d. by 1 m) within the 30.5 m regular grid (N=55). ‘ Variable Mean SD range CV Co C, C,,/(C0 + C.) R Model IGF (%) (%) (m) Bt 47.0 16.6 23-90 35 210.00 150 58 170 S 6.26-03 AWG-Ap 4 2 1-10 49 4.752 8 PN 7.3e-03 AWV-Ap 7 3 1-16 47 10.34 8 PN 9.2e-03 Ap-1500 10 2 6-16 22 2.183 3.431 39 110.0 S 6.56-03 Ap 26.7 3.2 18-38 12 9.400 8 PN 3.56-02 TC 1.29 0.36 08-233 28 0.067 0.0004 99 1.0 L 8.8e-03 P-clay 1 5 4 6-25 3 1 2 1 8 PN 2.96-02 P-clay+silt 36 10 16-58 29 1 10 8 PN 2.06-02 PAW 15 4 9-26 29 9.349 10.199 42 1 10.5 S 2.9e-02 WT-MAX 1.30 0.16 0.94-1.68 12 0.009 0.023 30 217.5 S 8.6e—03 WT-MIN 1.14 0.26 0.30-1.52 23 0.034 0.044 44 195.0 S 6.66-03 *Descriptions of symbols are presented in Table 1.2. 31 Table 1.4. Descriptive statistics and parameters for variogram models of variables obtained from soil cores within 30.5 m regular grid (N=55). ’ Variable Mean SD range CV CO C, CO/(C0 + C.) R Model IGF (%) (%) (In) Depth 0 to 7.6 cm BD 1.52 0.07 1 .32-1.67 5 0.005 a a PN 2.5e-03 TP 36 2 32-41 6 4.232 a 8 PN 1.2c-02 GM-6 l6 3 1 1-23 18 7.52 a 8 PN 5.06-03 GM—33 15 2 1 1-21 17 5.44 a 8 PN 2.86-03 GM-lOO 14 3 10-21 18 6.006 a 8 PN 5.06-03 VM-6 24 4 18-33 16 12.959 2.880 82 1 18.8 S 1.56-02 VM-33 22 3 17-30 15 7.439 4.199 64 91.2 G 6.28-03 VM-100 21 3 16-30 16 6.48 4.919 57 129.6 S 5.8e-03 APP-6 32 1 1 652 34 68.4 56.400 55 81 .6 S 1.06-02 APP-33 37 9 19-55 24 29.967 52.650 36 74.4 S 5.7e-03 APP-100 40 9 23-57 21 28.467 46.717 38 79.2 S 5.1e-03 Depth 7.6 to 15.2 cm BD 1.64 0.06 1.48-1.76 4 0.003 0.0008 81 58.8 S 1.8e-02 TP 35 2 32-40 6 4.80 a 3 PN 1 .3e-02 GM-6 15 2 1 1-20 17 4.992 1.8560 73 93.1 S 3 98-03 GM-33 l3 2 10-19 17 3.432 1.7670 66 97.6 S 1.3e-03 GM-100 13 2 9-18 17 3.016 2.1320 59 89.6 S 1.26-03 VM-6 24 4 18-33 16 8.26 6.4400 56 95.0 S l. le-03 VM-33 22 3 17-31 15 5.759 5.75 80 50 104.0 S 5.06-04 VM-100 21 3 16-31 16 4.509 6.4890 41 99.2 S 9.7e-04 AFP-6 31 9 10-49 29 45.044 48.445 48 104.0 S 2.16-03 AFP-33 38 8 22-54 21 28.975 35.907 45 1 10.4 S 2.26-04 AFP-100 41 8 24-55 19 30.00 33 48 124.8 S 8.96-04 2Descriptions of symbols are presented in Table 1.2. 32 A 1 4747000 4747850~ Northing (m) 4,. 4747800 8’ I. ;. qafigfiafiéb N A 4747000 \ 6 x . ,5 0:}R 4. J \\\\'\0\\ v‘ ‘ \04’0‘ .\\\\\\\k \.. . . Easting (m) 1 4 O O. 524/ ‘\\\\\\\\\s\ \ 0 0 f'ébl“§§ ‘93! ‘Qfifi§00w ‘b‘fib""‘§|le l r 0.0 §\\\O'\O;O;O‘: \\ ‘wnnsSSRfig .4 0 00» ‘310 0% 0 0 00 9’0 ’. I. r ’0 .0 d! V v {'3‘ 00' .0 O 30 47478505 , Northing (m) 47478007' S 0. .mgs‘. 257400 2 4747900 1 \ 5 2 257500 Easting (m) 257550 ,0 0 400.000%%&P’ \ 0 ; 0000 0 000' ‘49 00000000400000 . .\‘§§?A£§1§5§§%5fi§§§§0 . ' / ’ : 4 ”4%\\ 257600 A 257650 47478CD~I E . :3, 4747850 \ \ I: \\ . \\ ‘\<\\\\\V\ \\‘\ ..‘\\‘\\ 400 Easting (m) Figure 1.13. Interpolated maps of volumetric water content (%) at -6 kPa (A). -33 kPa (B), and -100 kPa (C) on surface 7.6 cm for the 30.5 m regular grid. Refer to table 1.1 for map unit descriptions. Northing (m) 34 l ‘ . _ :3 z ‘ ‘ \ . “:éififizzi‘ {tag 3?}ng ~ ”hfififi . $3. an?‘ gfi‘. gxfigg m k %§‘ i W //////' \.\ m¢%%%%%%z 1.2 :V‘ xfi ~ ' ;\ ‘ ggfifiigfiigi y \k t‘gxni /////// 50 25755 A Easting (m) Northing (m) Northing (m) //\// \ \ .>/\\\\ my}; a?” / \\\\\/\ / g. / i \\ ax \ / \ \ % //- 257400 257450 257500 257550 257600 257650 Easting (m) Figure 1.14. Interpolated maps of air filled porosity at -6 kPa (A), -33 kPa (B), and -100 kPa (C) for the 30.5 m regular grid. Refer to Table 1.1 for map unit descriptions. \ §Q§\ ‘mJ 35 7/ W, ', kg. //////// f7 " Northing (m) Easting (m) H I, , / 77 /l//%/%/ / My? ././//,/4y/ / 4,;//////\\X\\\\<%/%§MM\MZ 23 N A 257400 257450 257500 257550 257600 257650 Easting (m) B 4747900 "ill I... § " a 20 .E. Z - g 474755: Z " Z is z .- : 3E3: ' z .7. 5 ,_ . W \ - . \ . x t I. . \ \ I 1. . "z“ \ \’ // v ,. s” w... . . . . , _ ss- f . ’ rymwgggt . \ / 24/4 0 .5 .5 O N O 257500 257550 257600 257650 Easting (m) a" u : 'n':. 2. J- '5 257400 257450 Figure 1.16. Interpolated maps of Thickness of the Ap horizon (cm) (A), and total carbon (%) in the top 5 cm (8) for the 30.5 m regular grid. Refer to Table 1.1 for map units descriptions. E a 5 E O z ‘\ \.\\\ ~\\\‘» ‘ r. 257450 257500 257550 257600 A Easting (m) s§§s \\\,\\\ / " \.\\ . \ ’ .\\\\ Northing (m) A \l A N O 4747-u W\\\§t\ \ng \\ \ _ \\ \\\\\\\\\\§\ . 47478'| Z . /////;/2/%/z/ , 257400 257450 257500 257550 257600 257650 Easting (m) Figure 1.17. interpolated maps of profile clay content up to 1 m (%) (A), profile clay + silt content up to 1 m (%) (B). and profile available water (%) (C) for the 30.5 m regular grid. Refer to Table 1.1 for map unit desscriptions. 38 E 9.2.32 Easting (m) “"9 (m) Eas (B) water table depth (m) ' (A) and minimum xrmum Interpolated maps of ma for the 30.5 m regular grid in 1995. Refer to Table 1.1 for map unit descriptions. 1.18. Figure E; . ‘ \’/ V)’ 'X" ‘ l! i \I \ \\\‘;K\ mg ‘1'; .‘N 1 -'. {a} ~ .z. tits. 1 . ,3 iii . ’\ Northing (m) i. / fiQNse fly. \\ /\\\ ‘iisi «35'- ifl' 5%? ‘3 i \\%§§§lii‘i i . . 2' .l“1:.' , ii'pzz‘flxii“? ".4.‘ 3,: €.l§‘.. airmail; 1* . i :iizfiéli: Willi -. it. ~. . \ /////// /% / // // ’ / /, /////// 7/9///«/ - ZéO/i/ 36 \\\\\\ \\ \\\\\\\\ Northing (m) E 9: a. Easting (m) Figure 1.19. Interpolated maps of bulk density (g/cm3) (A) and total porosity (%) (B) on surface 7.6 cm for the 30.5 m regular grid. Refer to Table 1.1 for map units descriptions. 40 Variation in Surface Soil Properties Although bulk density was lower in the surface 7.6 cm than in the 7.6 to 15.2 cm depth, total porosity and water content and air-filled porosity at all measured matric potentials were similar between depths (Table 1.4). The CVs (%) for bulk density and total porosity were very low, in the mid-teens for water content, and ranged from 19 to 34 % for air-filled porosity, with highest CVs for the -6 kPa matric potential. Carter (1995) also reported low CVs (< 10 %) for soil bulk density and high variability at high matric potential. Mallants et a1. (1996) found similar results and suggested that at lower matric potentials, water is released by more uniform pore sizes These data were tested for normality (SAS, 1990) and no data transformation was needed for geostatistical analysis. Semivariograms were computed for each parameter. To examine for anisotropy, four direction-dependent semivariograms were calculated with lags grouped in 45° classes (0°, 45°, 90° and 135°). In no case was appreciable anisotropy evident; thus we assumed variograms to be isotropic and used omni-directional variograms (90° angular tolerance, i.e., direction independent) for the remainder of the analysis. In order to achieve acceptable precision and to smooth the structure of the variograms, maximum lag spacing equal to grid spacing, and lag tolerance equal to half the lag spacing were chosen (Issaks and Srivastava, 1989). All variograms were estimated using a minimum total of 1485 sample pairs, and a minimum of 214 sample pairs for each lag distance. Parameters of omni—directional semivariograms were determined based on the best-fitted model for soil properties at both soil depths. Indicative Goodness of Fit (IGF), a calculation performed by the program used for the spatial data analysis (V ariowin 2.2, 41 Parmatier, 1996), was used to quantify the traditional visual fit. The IGF is a number without units: a value close to zero indicates a good fit. Since it is a standardized measure of fit, its value is comparable from one modeling session to another, allowing a numerical check how well each model fits the experimental measures (Pannatier, 1996). Total porosity at both depths and bulk density and gravimetric moisture for all matric potentials for the surface 7.6 cm showed no spatial dependence as the semivariograms exhibited pure nugget effect (Table 1.4). Most other semivariograms were best described by the spherical model, except for a fit of the gaussian model for the volumetric moisture content at -33 kPa in the surface 7.6 cm. Except for a range of 59 m for the bulk density of the 7.6 to 15.2 cm layer, the range for most parameters centered on 1003: 25 m. Increasing the detail of sampling can reveal spatial structure in the apparently random effects of the pure nugget variances (Burrough, 1983 cited by Trangmar et al., 1985) and this principle will be addressed later. The ratio of the nugget effect to the sill (total semi-variance) enables classification and comparison among soil properties (Trangmar et al., 1985). This ratio was used to define distinct classes of spatial dependence for the soil variables as follows: if the ratio was 5 25%, the variable was considered strongly spatially dependent; if the ratio was between 25 and 75%, the variable was considered moderately spatially dependent; and if the ratio was > 75%, the variable was considered weakly spatially dependent (Chien et al., 1997; Cambardella et al., 1994). Moderate spatial dependence was found for most variables, except for volumetric moisture at -6 kPa at O to 7.6 cm depth and for bulk density at 7.6 to 15.2 cm, which were weakly spatially dependent. Patterns of moderate spatial dependency for soil properties were also found in studies by Cambardella et al. 42 (1994) and Chien et a1. (1997). Both intrinsic factors of soil formation, and extrinsic factors, such as soil management practices, may control spatial variability of soil properties. Usually, strong spatial dependence can be attributed to intrinsic and weak spatial dependence to extrinsic factors (Cambardella et al., 1994). In general, combination of both, results in soil parameters exhibiting moderate spatial dependence (Cambardella et al., 1994). Variation in Soil Profile Properties Table 1.3 lists the descriptive statistics and the results of the geostatistical analysis for water table depth measurements and for general physical properties obtained from undisturbed soil profile cores sampled at each grid point to a depth of l m. In general, these parameters showed a higher CV. in comparison with the data previously discussed (Table 1.4). No spatial dependence was observed for gravimetric and volumetric available soil water content of the Ap horizon, thickness of the Ap horizon, and profile clay and clay + silt content. The serrrivariograms were best described by the spherical model for the Ap horizon gravimetric water soil content at -1500 kPa, maximum and minimum water table depth, depth of the Bt horizon, and profile available soil water. A linear model was found to best fit the experimental data for total carbon content in the top 5 cm. This implies that the model is linear and does not reach a sill (Issaks and Srivastava, 1989). The range of spatial correlation varied from 110 m for the Ap horizon gravimetric soil water content at -1500 kPa to 217.5 m for maximum water table depth. The ratio of nugget variance to total semi-variance was 39 to 58%, indicating moderate spatial dependence. 43 Analysis of soil variability using semi-variograms has aided identification of soil mapping units and placement of mapping unit boundaries (Webster, 1985). Trangmar et al. (1985) presented several examples where semi-variance analysis was used to quantify spatial dependence of soil genetic processes and soil-forming factors such as rainfall, parent material composition and deposition. At our site, no spatial dependence was found for thickness of the Ap horizon and profile clay and clay + silt content, parameters that are traditionally used for defining the boundaries of soil map units. Lack of spatial dependence, also, indicates that the area may contain considerable inclusions that are important sources of variability and would not be shown in soil survey maps. Kriging techniques were used to make contour maps of all soil parameters. Kriging is a local estimation technique which provides the best linear unbiased estimation of variables at unsampled locations using the structural properties of semi-variogram and the initial set of data values (Trangmar et al, 1985). For those variables that exhibited no spatial variability, the inverse distance squared was used as interpolation method. The information shown in contour maps is useful to gain a better understanding of the spatial distribution of soil properties and to visualize and define different management zones in a given area. Figures 1.12 to 1.19 show the soil map units overlaid by the interpolated soil property maps generated from the 30.5 m regular grid, in order to determine the extent of alignment between the two maps. Not a single soil property aligned with soil map mrits (Figures 1.12 to 1.19), which confirms that the soil map units did not reflect the soil variability present within this particular field. However, the maps generated 44 geostatistically showed quite well that the experimental area displayed a sandier Ap horizon in its central western part and a clayey Ap horizon in its eastern part. In general, clayey areas coincided with a relatively higher water table (Figure 1.18), lower elevation (Figure 1.1b), higher volumetric water content (Figure 1.13), shallower Bt horizon (Figure 1.10c), and higher total carbon in the t0p 5 cm (Figure 1.16b). These observations suggest that the spatial coincidence of these properties could form the basis for defining management zones. Effect of Scale on Estimation of Spatial Variability in Surface Soil Properties Soil physical prOperties measured from intact soil cores obtained from the 0 to 7.6 cm depth were similar for the 30.5 m grid (N=55) and the geospatial grid (N=l62), although air-filled porosity values were higher and less variable (lower CVs) for the more intensive sample values (Tables 1.4 and 1.5). The semivariogram for the geospatial grid values for bulk density, total porosity, and the measured gravimetric water contents showed strong spatial dependence in these soil properties that was not evident in the semivariograms for the 30.5 m grid values. The semivariograms for volumetric water contents and air-filled porosity for the geospatial grid values showed lower nugget effect (C,) and structural (C,) variances, lower ratios of the nugget effect to the sill, and smaller ranges than those derived from the 30.5m grid values. The best fit model of the semivariograms were primarily exponential for the geospatial grid and spherical for the 30.5 m grid. While descriptive statistics for the two sampling schemes are very similar, the spatial distribution of the soil physical properties predicted by the kriging the semivariograms will be quite different as evidenced by maps for the volumetric water 45 contents at -33 kPa matric potential (VM-33) obtained for the two grid scales in Figure 1.20. The interpretations relative to site-specific management based on these maps would be quite different. Table 1.5. Descriptive statistics and parameters for variogram models of variables obtained fiom soil cores (7.6 by 7.6 cm) within the geostatistical grid (N=162).‘ Variable Mean SD range CV Co C, Ca/(Co + C.) R Model IGF (%) (%) (m) BD 1.52 0.08 1.30-1.67 5 0.003 0.0032 50 17.9 S 1.6e-03 TP 36 2 31-44 7 2.006 4.189 32 13.3 S 4.9e-03 GM-6 14 2 2-20 15 2.927 2.543 54 68.6 S 1.1e-03 GM-33 13 2 9-19 14 0.296 4.0 7 46.2 E 5.6e-O3 GM-IOO 12 2 9-18 15 0.828 3.276 20 45.5 S 1.3e-02 VM-6 22 3 3-28 14 5.369 4.732 53 66.5 S 2.4e-02 VM-33 20 2 14-27 13 0.462 7.5 6 58.8 E 8.9e-03 VM-lOO 19 2 13-26 14 0.395 7.5 5 53.2 E 8.8e-03 MP6 39 8 21-89 21 41.89 23.43 64 9.0 S 3.8e-02 AFP-33 44 7 27-62 16 11.128 46.106 19 43.4 E 5.4e-O3 APP-100 46 7 29-64 16 12.72 46.106 22 47.6 E 5.0e-03 ’ Descriptions of symbols are presented in Table 1.2. 9” ‘ ”’///////////// v / /7/7/ // 47/ / ’ / // / //// //,// /// , . , . //'2%/“ , //// /////// /fi/ E 47.78.. / // /\ gfiggtggfig . 27 N A 257400 257450 I I a'gm 47850— m; s s“ \\\\\ 47800- \\\\ / / 2 / , 400 57'500 257550 00 Easting (m) Figure 1.20. Volumetric water content (%) at -33 kPa on surafce 7.6 cm for the 30.5 m grid scale (A) and the geospatial scale (B). Refer to Table 1.1 for map unit descriptions. Northing (m) / , >\\ \\/// / i\\\\ , / \\ 257500 257550 Northing (m) \ Easting (m) Figure 1.21. Interpolated map of clay (A), silt (B), and sand (C) contents (%) on surface 18 cm for the geopatial grid (N=419). 48 Table 1.6. Descriptive statistics and parameters for variogram models of soil textural analysis of the Ap horizon for the 30.5 m regular grid (N=55) and for the geostatistical grid (N=419). Variable Mean SD range CV Co C, C.,/(Co + C.) R Model IGF (%) (%) (%) ("1) Regular grid Clay 1 1 3 6-18 26 5.524 5.439 50 161.5 G 5.6e-03 Sand 65 7 45-79 1 1 36.18 26.456 58 98.9 G 1.6e-02 Silt 23 6 12-38 24 13.437 22.079 38 103.4 S 9.2e-03 Geostatistical grid Clay 1 1 3 6-27 24 3 .477 6.033 36 175 S 1.9e-02 Sand 67 7 45-82 10 4.4 43.12 9 52.7 S 1.0e-02 Silt 22 5 5-38 23 3.778 24.57 13 52.7 S 3.8e-03 *Descriptions of symbols are presented in Table 1.2. Soil texture (sand, silt, and clay) measured from the geospatial grid (N=419) and the 30.5 m grid had similar descriptive statistics but much stronger spatial dependence was detected in the geospatial grid values (Table 1.6). The nugget (C,) variances were lower in the semivariograms determined from the geospatial grid than those determined from the 30.5 m grid but the structural (C,) variances were higher, creating much lower ratios of the nugget effect to the sill (9 to 36% versus 50 to 58%). The range for clay was similar for geospatial and 30.5 m grid sampling schemes (161 m versus 175 m, respectively) whereas the range was about 50% lower in the semivariograms of the geospatial grid for sand (52.7 versus 98.9 m) and silt (53 versus 103 m). The best fit model was spherical for all textures in the semivariogram from the geospatial grid and gaussian for clay and sand for the semivariogram from the 30.5 m grid. Kriging the semivariograms produced quite different soil texture maps for the experimental area (Figure 1.9 and 1.21) The 30.5 m grid soil texture map showed less variability or a 49 smoother contour map, mainly for clay and sand content, where the best fit model was gaussian (Figure 1.9a and 1.9c). Similar patterns in those maps were not very well defined if compared to the geospatial grid. Although the silt content map for the geospatial grid (Figure 1.21b) showed more variability, it is possible to visualize similar patters in the central eastern part, compared to the silt content map for the 30.5 m grid (Figure 1.9b). In general, similar patterns when present, were very poor defined for all texture maps. When soils were sampled in detail using geostatistical designs, soil texture and soil physical properties exhibited strong spatial dependence allowing for precise soil property maps for the experimental area. The 30.5 m grid proved inadequate in detail and/or sample spacing for estimating the semivariogram for the measured soil physical properties and would not be a suitable sampling scheme for creating accurate soil property maps. The cost of soil sampling and analysis preclude these detailed studies on the farm. However, sensor technologies may allow mapping soil physical properties at the required intensity or allow direct mapping of soil management zones fi'om sensor readings (Sudduth et al., 1997). The coarse grid sampling schemes, however, provide good estimates of descriptive statistics for soil physical properties and may be quite useful in directed sampling schemes for estimating average properties of predetermined site-specific management zones. Semi-variance analysis and kriging interpolation techniques demonstrated that there were similarities in patterns for some of the soil parameters. This suggests that areas exhibiting similar patterns could be defined as different management zones. However, because the spatial variability of soil physical properties is strongly influenced by the 50 scale of the investigation, it remains to be seen whether or not these results will be useful for extrapolating spatial information obtained at the field scale to the watershed or regional scale. The results showed that sampling efficiency is a function of spatial dependence of the variable of interest. As the distance of spatial correlation increases, fewer observation sites are needed without significant information loss. However, before a particular sampling design scheme is selected, one would have to consider cost effectiveness of the available scheme, as well as practical considerations such as time constraints, site accessibility, cost of the laboratory analyses for each sample, among others site specific issues. CONCLUSION The soils within the experimental area were quite variable and did not correspond well to the soil map units given in the county soil survey maps. The soil survey for this field is not accurate and of little value to SSM. The 30.5 grid was descriptive but did not accurately describe the spatial variability in soil physical properties. The intensive geostatistical sampling designs produced very good estimates of the semivariogram and indicated a strong spatial dependence of all soil physical and soil texture properties measured. The cost of intensive soil sampling, even at the 30.5 m grid spacing, are cost prohibitive to any practical application in SSM. This fact points to sensors of soil properties as needed to make assessment of soil physical condition reasonable and efficient. CHAPTER 2 Spatial Variability in Soil Physical Properties and Corn Yield Induced by Tillage INTRODUCTION The major focus of site-specific management (SSM) or precision farming has been nutrient management; however, variability in crop yield has not generally corresponded to variation in nutrient availability or to the variable rate application of fertilizers (Everett and Pierce, 1996). Water availability to plants is a major factor- regulating crop yields but is often not been a major consideration in SSM for agriculture, in part because water availability is weather dependent and is difficult and/or expensive to measure in space and time. Within a given climate, the availability of water can be related to soil physical and hydraulic properties, water table, and landscape effects, which vary spatially and temporally. Studies on crop yield variability have cited higher yields in lower landscape positions where soil water availability was higher (Spomer and Piest, 1982; Stone et a1. 1985; Hanna et al., 1982). In southeastern Washington, higher yields in lower landscape positions, such as on interfluves and toeslopes, were attributed to differences in surface soil thickness (Ciha, 1984). Combinations of soil properties and landscape geometry can account for yield variability. Khakural et al. (1996) found that depth to free CaCO3, surface available P, available K, relative elevation, and slope gradient explained 65% of the variability in corn yield while depth to free CaCO3, surface available P, relative 51 52 elevation, slope gradient, and profile curvature explained 78% of the variability in soybean yield. Few examples exist, however, where soil physical and hydraulic properties have been used to explain crop yield variability. Tillage is a major agricultural input affecting soil physical, chemical and biological properties with impacts primarily at or near the soil surface (Blevins, et al., 1983; Hill, and Meza-Montalvo, 1990; Seta et al., 1993; Kitur et al., 1993; Hubbard et al., 1994; Richardson and King, 1995). Management activities like plowing and leveling are known to influence the spatial variability of structure-related soil properties (Bouma and Finke, 1993). F inke et a1. (1992), for example, found that the thickness of a disturbed soil displayed a spatial structure that reflected the surface topography before the leveling took place. Hydraulic properties of the surface soil layer affect water through their effects on infiltration, evaporation, soil water storage, and conductivity of water to and away from the soil surface. Tillage and associated crop residue management also affect runoff and erosion, further confounding these processes (Seta et al., 1993; Richardson and King, 1995). The extent to which tillage and crop residue management alter the spatial variation in soil properties and subsequently crop yield are of interest in SSM because tillage type and intensity and the timing of tillage could be altered site-specifically (V oorhees et al., 1993) Site-specific tillage may be desirable since soils vary with regard to optimal tillage needs for crop production and for conservation of soil and water (V oorhees et al., 1993) and many different soil and conservation needs can be present within a fields. The practicality of site-specific tillage will vary by location and tillage type. For example, in rolling landscapes, where many soils are in close association ranging from upland to 53 depressional in nature, it may be not practicable to till the soils as separated entities although the intensity of tillage could be varied. When soil series are more distinct, then individually designed tillage systems can be developed for specific soil series or landscape position (V oorhees et al., 1993). Khakural et al. (1992) measured soil properties across two discrete but closely associated landscape positions and found that the improvement in soil properties on an upland soil with conservation tillage out- weighed the effects of these same high residue tillage on a depressional soil. Managing tillage operations according to spatially varying soil characteristics has the challenge of trying to satisfy multiple, and often opposing objectives, e.g., soil conditions best for plant growth may not be best for erosion concern or pollution impact. Because water availability is so important to crop yield and because tillage impacts important soil properties and processes that regulate water availability either directly or indirectly, research on managing tillage operations according to spatially varying soil characteristics remains a important need. This study assesses the co-spatial variability of soil physical properties and corn yield under long-term no-tillage and how tillage affects those spatial relationships for a glacially derived landscape in central Michigan. MATERIAL AND METHODS A field study was conducted in 1995 and 1996 in a 3.72 ha section of a larger field located 6 km south of Durand, Michigan (47° 47’ 30”N, 83° 52’ 30”W). The field had been managed in a com-soybean rotation under no-tillage for more than 10 years. Paired strips 4.57 m wide and 305 m long were established across the experimental area in April 54 1995 to evaluate tillage effects on the spatial variability of soil properties and corn grain yield. Chisel plowing was assigned randomly to a strip within each of 10 replicated pairs (Figure 2.1) creating a randomized complete block experimental design with 10 replications. Chisel plowing followed by one pass of a field cultivator was performed in the spring on 3 May, 1995 and 6 May, 1996. In both years, corn was planted on 8 May at 66000 plants ha" (hybrid Pionner brand 3733). In 1995, no starter fertilizer was applied, while in 1996 consisted of 94 L ha'1 of 10-34-00 (N -P-K). Nitrogen fertilizer applications consisted of 145.6 kg N ha’1 on 3 July, 1995 and 150 kg N ha"on 27 July, 1996. Weed control followed standard recommendations for the pre-emergence applications for this area for corn. Corn grain yields for both tillage systems were obtained every 15 m along the length of each strip from the center two rows of the 6-row plots. Plant populations at harvest were measured at the same area. At each 30.5 m grid point where soil profiles had been described (Figure 1.3, Chapter 1), 10 consecutive plants within 15 m section of two corn rows centered on grid intersection were harvested and oven-dried to determine stover yield. 55 :1 Chisel plow [:1 No-tillage Figure 2.1. Location of the grid soil profile samples relative to the map units from the Order two soil survey (A) and detail of the tillage strips (B). Refer to Table 1.1 for map unit descriptions. 56 At the 30.5 m regular grid locations, water table depth was measured weekly throughout the growing season using piezometers installed to a depth of 1.5 m. Volumetric soil water content in the surface 18 cm was measured by time domain reflectometry (TDR) in both tillage systems in the proximity of each grid point. Soil profile volumetric water content (soil profile storage water) was monitored weekly at every other 30.5 m grid point by neutron probe access tubes, 1.5 m deep, installed in both tillage systems. Water table depth, surface volumetric soil water, and profile volumetric water content were always measured on the same day. Undisturbed soil cores were obtained within the same 100 by 100 sub-area of the experimental site as described in Chapter 1 for the chisel plow and no-till areas. Soil hydraulic properties (volumetric soil water content; pore volumes at -6, -33, -100 kPa matric potential, total porosity; and bulk density) were measured as described in Chapter 1. Soil samples were obtained from each tillage treatment at each 30.5 m grid point from the 0 to 5 cm depth on 15 October 1997. Total carbon was determined by dry combustion using a CHN analyzer (Carlo Erba Instruments, Italy). and wet aggregate stability by the single sieve method with correction for the presence of primary particles: sand and coarse fiagments (Kemper and Rosenau, 1986). Tillage effects on crop yield for the whole field were evaluated using analysis of variance for a randomized complete block design with 10 replications. Tillage means for crop yield were tested using LSD at 0.05 probability level. A t-test for paired samples was used to test the effect of tillage on total carbon, aggregate stability, average seasonal volumetric water content, average seasonal profile water content obtained at each grid point. 57 Geostatistical analyses were performed as described in Chapter 1 for soil and plant variables for both tillage systems. Geostatistical methods were used to measure and model the spatial correlation for soil parameters and corn yields for both tillage systems. Variograms were derived by plotting the semivariance of the soil and plant data as a function of vector distance. Each parameter was subjected to semivariance analysis using isotropic and anisotropic models (linear, spherical, exponential, and gaussian) for defining semivariograms. The models were then used along with kriging techniques to develop maps showing spatial patterns in variability of selected soil properties, corn yield and plant variables (Issaks and Srivastava, 1989). Descriptive statistics and correlation were calculated using SAS (SAS, 1990). Stepwise regression analysis was used to model the effects of soil physical properties on corn yield. The variables and respective symbols used in the regression analysis are listed in the Table 2.1. 58 Table 2.1. Description of variables with the respective symbols used in the stepwise regressions. Symbol Description AF P6 Air filled porosity (%) at -6 kPa matric potential as described in Chapter 1. AF P33 Air filled porosity (%) at -33 kPa matric potential as described in Chapter 1. AF P100 Air filled porosity (%) at -100 kPa matric potential as described in Chapter 1. Ap Thickness of the Ap horizon (cm); Ap-1500 Gravimetric moisture content at -1500 kPa matric potential (%) of the Ap horizon Ap-clay Clay content of the Ap horizon (%) Ap-silt Silt content of the Ap horizon (%) Ap sand Sand content of the Ap horizon (%) BD Bulk density (Mg m”) Bt Depth to the Bt horizon (cm); G6 Gravimetric water content (%) at -6 kPa matric potential as described in Chapter 1. G33 Gravimetric water content (%) at -33 kPa matric potential as described in Chapter 1. G100 Gravimetric water content (%) at -100 kPa matric potential as described in Chapter 1. PAW Weighted average profile available water (%) P-clay Weighted average profile clay content (%) P-clay+silt Weighted average profile clay + silt content (%) POP Plant population (plants ha") TC Total carbon content in the top 5 cm (%) TP Total porosity (%) V6 Volumetric water content (%) at -6 kPa matric potential as described in Chapter 1. V33 Volumetric water content (%) at -33 kPa matric potential as described in Chapter 1. V100 Volumetric water content (%) at -100 kPa matric potential as described in Chapter 1. VWC-tasseling Volumetric water content (%) at the top 18 cm measured by TDR during the corn tasseling period. VWC-silking Volumetric water content (%) at the top 18 cm measured by TDR during the corn silking period VWC-AVE Average volrunetric water content (%) at the top 18 cm measured by TDR during the growing season. 59 Table 2.1 (cont’d) VWC-MAX VWC-MIN WT-MIN WTuMAX WT-AVE WT-tasseling WT-silking Maximrun volumetric water content (%) at the top 18 cm measured by TDR during the growing season. Minimum volumetric water content (%) at the top 18 cm measured by TDR during the growing season. Minimum depth (m) of the water table during the growing season. Maximum depth (m) of the water table during the growing season. Average depth (m) of the water table during the growing season Water table depth (m) during the corn tasseling period. Water table depth (m) during the corn tasseling period 60 RESULTS AND DISCUSSION Three important aspects of this research are addressed in sequence as follows: (i) how tillage, after more than a decade of no-tillage management, altered soil physical properties spatially; (ii) how corn yield varied spatially over a two year period under different tillage management; and (iii) the extent to which soil physical properties, water table, and soil water content during the growing season explain the spatial variability in corn yield for the two years of this study; and whether these differences related in any way to differences in corn yield in the two tillage systems. Tillage Effects on Soil Properties The surface 5 cm of soil in the chisel plowed areas had, on the average, 22% lower total carbon (TC) (11.9 g kg“) than soil managed under no-tillage (15.2 g kg") (Table 2.2). The Host was significant for differences in TC due to tillage (Table 2.3). Table 2.2. Descriptive statistics and variogram model parameters of total carbon (TC) and aggregate stability (AS) on surface 5 cm for both tillage systems after two growing seasons (1996).: Tillage Variable Mean range CV C, C, C,/(C,+ C,) R M1 IGF (%) (%) (m) TC (g kg") NT TC 15.2 7.7-27.9 29 43.196 128.396 34 104 S 4.2e-02 CP TC 11.9 7.0-20.5 32 45.6 133.196 34 102 S 4.2e-02 AS (%) NT A8 82.8 47.1-97.4 11 363.97 1019.14 36 102 S 4.1e-02 CP AS 72.5 44.3-95.3 20 191.67 773.852 25 87 S 3.2e-02 IM = Model 3Descriptions of symbols are presented in Table 1.2. 61 Table 2.3. Results of paired t statistic test on means of selected soil properties for different tillage systems after two growing seasons (1996). TC AS. VWC PW? Mean 3.3 10.3 0.026 0.014 STD error 0.04 2.29 0.002 0.0079 L0 95% CI 2.6 5.62 0.021 -0.0025 UP 95% CI 4.2 14.98 0.030 0.314 T 8.33 4.48 12.36 1.83 P 0.0000 0.0001 0.0000 0.0892 I Number of observations N=15. TC = Total carbon (g kg"); AS = Aggregate stability (%); VWC = volumetric water content (m3 m3) at the top 18 cm measured by TDR. PWC = average profile volumetric water content (m3 m”) up to l m depth measured by neutron probe. L0 95% CI = lower 95% confidence interval. UP 95% CI = upper 95% confidence interval. STD error = standard error. T = test statistic t P = p-value. Small p-value means that the mean of the difference is not zero. The p-value is for a two tailed test. 62 The spatial dependence of TC was moderately strong for both tillage systems and so was the difference in TC between tillage systems (Table 2.2). These differences in TC may result from carbon losses due to tillage as indicated in other tillage studies (Elliott et al., 1994; Gupta et al., 1994; Guillermo et al., 1997) but may also reflect a movement of carbon accumulated in the surface of no-till being moved to a lower depth with tillage. This would be consistent with reports on the effect of tillage the positioning of carbon within the soil (Pierce et al., 1994; Staricka et al., 1991). Average aggregate stability (AS) in the chisel plow soil (72.5%) was 10% lower than the no-till soil (82.8%) (Table 2.2). For a small percent of the sample grid points, AS was higher in the chisel plow than in the corresponding no-till soil. This may reflect small-scale variability in AS that would not be accounted for in the sampling scheme since tillage locations were not exact. In any event, the t-test was significant for differences in AS due to tillage (Table 2.3). The spatial dependence of AS was moderately strong for both tillage systems but the difference in AS between tillage systems was weak, largely due to the occasions where the difference between tillage systems reversed direction (Table 2.2). Tillage, as would be expected, generally reduced aggregate stability and is consistent with other studies (Guillermo et al., 1997; Caron et al., 1996). Tillage, as expected, altered the soil physical properties determined on intact soil cores (Table 2.4). While average values for the soil properties for each tillage system was within one standard deviation of the means, tillage eliminated (pure nugget effect of the semivariogram) or reduced (higher nugget effect to total variance ratio) the spatial dependence of soil physical properties established under long-term no-tillage. 63 Table 2.4. Descriptive statistics and variogram models parameters of variables obtained from soil cores (7.6 by 7.6 cm) within the field geostatistical grid (n=l62) for both tillage systems.1 Variable Mean SD range CV C, C, C,/(C, + C,) R Model IGF (%) (%) (m) No-till BD 1.52 0.08 1.30-1.67 5 0.003 0.003 50 17.9 S 1.6e-03 TP 36 2 31-44 7 2.006 4.189 32 13.3 S 4.9e-03 GM-6 14 2 2-20 15 2.927 2.543 54 68.6 S 1.1e-03 GM-33 13 2 9-19 14 0.296 4.000 7 46.2 E 5.6e-03 GM-lOO 12 2 9-18 15 0.828 3.276 20 45.5 S 1.3e-02 VM-6 22 3 3-28 14 5.369 4.732 53 66.5 S 2.4e-02 VM-33 20 2 14-27 13 0.462 7.500 6 58.8 E 8.9e-03 VM-lOO 19 2 13-26 14 0.395 7.500 5 53.2 E 8.8e-03 AF P-6 39 8 21-89 21 41.89 23.43 64 9.0 S 3.8e-02 AFP-33 44 7 27-62 16 1 1.13 46.1 1 19 43 .4 E 5.4e-03 APP-100 46 7 29-64 16 12.72 46.1 1 22 47.6 E 5.0e-03 Chisel plow BD 1.41 0.10 121-164 7 0.010 a a PN 3.3e-03 TP 39 3 32-47 7 7.395 1.652 82 56.6 S 5.4e-03 GM-6 15 2 10-21 14 2.385 2.789 46 52.4 G 9.5e-03 GM-33 13 2 10-18 14 0.998 3.293 23 60.7 S 8.3e-03 GM-lOO 13 2 9-17 14 0.917 3.161 22 57.3 S 1.3e-02 VM-6 21 3 15-30 13 4.514 3.996 53 54.5 G 3.5e-03 VM—33 19 2 l3-25 13 2.664 5.004 35 66.9 S 5.3e-03 VM-100 18 2 13-24 14 2.751 4.671 36 62.8 S 5.3e-03 AFP-6 46 8 24-61 17 61.44 a a PN 1.2e-02 APP-33 51 7 26.66 15 44.65 17.98 71 69 S 2.8e-03 APP-100 53 7 28-69 14 42.75 17.67 71 69 S 3.7e-03 * Descriptions of symbols are presented in Table 1.2. 54 Differences in spatial structure between tillage systems would indicate that the differences between tillage systems were spatially dependent. On the average, soil water retention, both gravimetric and volumetric, was not greatly affected by tillage and for the most part both tillage systems had similar spatial dependence. There was a change in the best fit model for the senrivariograrn for soil water retention and air filled porosity, and an increase in the ratio of nugget to total variance and range for most matric potentials. Changes due to tillage in these properties similar in magnitude were reported for the Capac Loam soil (fme-loamy, mixed, mesic Aerie Ochraqualf) in Michigan (Pierce et al., 1994). The CVs were < 10 % for bulk density and total porosity and between 13 and 21% for the other soil properties. The volumetric water content (V WC) in the surface 18 cm for the growing season averaged 0.026 m m'3 (Table 2.3) lower in chisel plow than in no-tillage as has been reported elsewhere (Hubbard et al., 1994; Griffith et al., 1986; Waddell and Wei], 1996; Azooz et al., 1996). Soil profile water content measured with a neutron probe on the average over the growing season showed no significant differences over the l m depth between tillage systems. The upper 75 cm portion of the soil profile, however, did show tillage effects, with higher volumetric water contents in no- till (Figure 2.2). 65 20 40 60 80 Depth (cm) 100 8/9/96 1 20 20 40 60 80 1 00 1 20 Depth (cm) 8/1 9/96 8128/96 20 40 60 80 1 00 1 20 Depth (cm) 9/1 1 I96 9/20/96 % % Figure 2.2. Percent difference (NT-CP) in profile volumetric water content at different depths and on different dates. 66 Tillage Impacts on Corn Yield Average corn grain yields were similar for 1995 and 1996 (8477 and 8869, respectively) but the standard deviation, range, and CV were higher in 1996 (Table 2.5). Corn grain yield in the chisel plow treatment averaged 298 kg ha‘I higher than no-till in 1995 (LSD = 228 kg ha") but average yields were not different in 1996. Starter fertilizer was not applied in 1995 and given the potential for starter fertilizer benefits in chisel plow, this may be a factor in this observation. On a field average, chisel plowing had a small yield advantage in 1995 and 1996; however, given costs of energy and equipment and changes in soil quality, the difference may not be economical. Of interest here was whether the spatial dependence of corn grain yield would change if the soil were plowed and under what condition corn grain yield would be different under different tillage management. Corn grain yields varied spatially each year, more so in 1996 than in 1995, and varied differently with tillage (Figure 2.3 and 2.4). Corn grain yields were spatially dependent in both years but yields in 1996 exhibited moderate to strong spatial dependence while spatial dependence of yield was weak in 1995, as evidenced by lower nugget effect to total variance ratios, (C,/(C, + C,); Table 2.5). The structural variance (C,) was 6 to 13 times higher in 1996 than in 1995. Nugget effect (C,) were similar between years within a tillage system but were higher in no-till than chisel plow both years. The two year average corn grain yield for each tillage system showed slightly higher average yield for chisel plow (8803 versus 8542 kg ha" for no-till) but similar variances and very similar spatial dependence for the two tillage systems (Table 2.5). 67 Table 2.5. Descriptive statistics and variogram model parameters of corn yield for both tillage systems in 1995 and 1996.1 Year Mean SD range CV C, C, C,/(C, + C,) R M IGF kg ha" Kg ha" (%) (%) (m) No-till (NT) 1995 8328 706.4 5826-9832 8 385000 100000 79 57.8 G 3.6e-03 1996 8757 1007 2653- 10535 1 1 384963 582963 40 81.6 E 1.5e-03 Average 8542 687.5 4584-9833 8 168000 264000 39 59.4 E 4.4e-O3 Chisel plow (CP) 1995 8626 542.4 6258-10096 6 228000 57000 80 49.3 S 1.3e-03 1996 8980 1078 20 19- 10799 12 264000 756000 26 56.1 E 5 .3c-03 Average 8803 652.7 5007-10077 7 124700 258000 32 59.5 E 4.8e-03 Difference NT-CP 1995 -298 715.2 -2320-1956 387566 112183 77 62.9 G 4.2e-03 1996 -223 886 -3299-2590 655700 71 100 90 170 S 5.7e-03 Average -261 594 -2346-1383 280740 53952 84 163 S 5.2e-03 * Descriptions of symbols are presented in Table 1.2. Northing (m) Northing (m) Northing (m) 68 257450 257500 257550 Easting (m) 1'.','"J""”\ I llllll 2° ,' 300 Ill... . l ‘ . 257400 257450 257500 257550 257500 0 Easting (m) |||||||"'®WI ‘|"" :5:- .'3°° \\\\\\\\\\\ \ll 3:: . //////\' llllll \<\. l l " lllllll l .. a * lllllll ||||"‘ "mill 2 llllllllll -900 // ////_ lllllllm ' 4747800 llllunllllllx\\\\,\>>§llll ..... ..... "Illl "" Illll" 5 257500 257550 Easting (m) 257400 257450 257 600 257 650 Figure 2.5. Interpolated maps of average of the two years of com grain yield (kg/ha) for no-tillage (A), chisel plow (B), the difference between no-tillage and chisel plow by subjecting the differenceto geostatistical analysis (C), and by subtracting (B) from (A) (D). E S 71 Where spatial dependence is moderate to strong, the exponential model fit the semivariogram for grain yield (1996 and two year average) while the model varied between tillage systems in 1995 where spatial dependence was week. The spatial differences in yield between tillage systems provides an opportunity to identify site-specific tillage zones. To be useful, difference maps should not vary substantially between years, or the average difference among many years must relate to some economically manageable zone delineation. Yield difference maps for a given year or multiple years can be determined by subtracting the maps created from kriging the semivariograms from the geostatistical analysis of the two tillage systems or by calculating the yield difference between the two tillage treatments at all yield locations and subjecting the difference to geostatistical analysis as was done for the tillage yields independently. Yield difference maps derived using both techniques for each year are given in Figure 2.3c,d and 2.4c,d and for the average of the two years in Figure 2.5 c,d. Calculated yield differences subjected to geostatistical analysis exhibited weak spatial dependence in both years and for the two year average, as evidenced by a high nugget variance to total variance ratio (77% for 1995, 90 % for 1996, and 84% for the two year average; Table 2.4). For 1995 and the two year average, the two difference maps are similar. Recall, however, that the large nugget variances for 1995 and the two year average indicate large prediction errors for these maps. For 1996, there is considerably more variability expressed in the subtracted map (Figure 2.4d) than the kriged difference map (Figure 2.4c). Since the subtracted map was generated from maps with high spatial dependence whereas the spatial dependence of the difference was weak, the subtracted map may be a more realistic representation of the real difference between the tillage 72 systems. In any case, the extent to which the average map delineates economically viable management zones is uncertain. Lamb et al. (1997) found a weak relationship among years between relative yield and location in a 1.8 ha field during 5 years of study in Minnesota. Therefore, it may take many years of yield map data to delineate management zones within a field. Corn grain moisture at harvest was similar for both tillage systems for 1995 and 1996 and variability was low (CVs ranged from 3 to 6 %; Table 2.6). Grain moisture exhibited strong spatial dependence in no-till and moderate spatial dependence in chisel plow in both (low to moderate nugget to total variance ratios; Table 2.6). The best fit models to semivariograms were spherical for no-tillage and gaussian for chisel plow. The range was higher in 1995 than in 1996 and slightly higher for no-till than chisel plow. Plant populations averaged near the target population of 66,000 plants ha", with slightly higher values in chisel plow than no-till and higher in 1996 than in 1995 (Table 2.6). Variation in plant population was similar to grain yield (Table 2.6), with CVs ranging from 6 to 11 % and generally moderate to weak spatial dependence, with no spatial dependence found for no-till in 1996. The model fit to the variogram and the range varied within years and tillage systems for plant populations. Stover biomass measured at harvest was quite variable (CVs ranged from 18 to 22%). In 1995, stover was higher in chisel plow than no-till but the reverse was true in 1996. Spatial dependence of stover was moderate to strong, depending on year and tillage system, and the spherical model fit the semivariogram in all cases, with the range varying from 88 to 110 m. 73 Table 2.6. Descriptive statistics and variogram models parameters of grain moisture (GrM), plant population (PP), and stover yield (SB) for both tillage systems in 1995 and 1996.3 Variable Mean SD range CV C, C, C,/(C,+ C,) R Model IGF (%) (%) (m) No-tillage - 1995 GrM 19 0.55 17.1-20.1 3 0.0599 0.3639 14 114 S 8.9e-03 PP 60716 6879 35254- 11 2.6e+07 2.6e+07 50 80 G 1.3e-02 73255 SB 3490 636 2463- 18 179189 3455786 52 105 S 4.3e-02 4906 No-tillage - 1996 GrM 19.7 1.15 16-22.5 6 0.1399 1.4881 9 68 S 7.1e-O3 PP 63701 6381 25633- 10 3.5e+07 a a 79188 SB 3999 794 1973- 20 1511720 466116 32 100 S 410-02 5659 Chisel plow - 1995 GrM 19.1 0.49 17.5-20.1 3 0.09 0.2639 34 92 G 5.6e-03 PP 63732 4143 50820- 6 1.2e+07 l.9e+07 63 92 G 1.8e-02 74172 SB 4358 950 2941- 22 866320 5506470 16 88 S 3.2e-02 8373 Chisel plow - 1996 GrM 19.8 1.10 16.6-23 6 0.5069 1.3649 37 59 G 8.5e-03 PP 66711 7114 16021- 11 3.0e+07 4.2e+07 71 29 S 6.4e-03 77815 SB 353 1 700 2578- 20 1946340 3595790 54 1 10 S 4.4e-02 5365 ‘ Descriptions of symbols are presented in Table 1.1. 74 Influence of Soil Properties on Corn Grain Yield For 1995, using stepwise regression analysis (PS 0.05), 52 % of the variability in corn yield was explained by a linear combination of volumetric water content in the surface 18 cm at tasseling (V MC-tasseling), at silking (V MC-silking), and the average in the soil surface during the growing season (V WC-average; Table 2.7). The negative coefficient on the VWC-average variable may reflect the negative impacts of excessive water during the wet conditions during the early portion of the growing season in 1995. Positive coefficients for VMC-tasseling and VMC-silking reflect the importance of water during the reproductive period for corn. For 1996, only one significant regression was observed; a simple linear regression of corn yield on clay content of the Ap horizon explained 25 % of the variability in yield. No other linear combinations of soil properties measured in Chapter 1 or water table depths or soil water contents measured during the growing season (regression variables given in Table 2.1) were significant in explaining yield variability in either year. Table 2.7. Results of stepwise regression of no-tillage corn grain yield and soil properties for two years. Year Variables Coefficient P-value l 995 Constant 10919.30 0.0000 VWC-tassiling 221.07 0.0082 VWC-silking 121.25 0.0444 VWC-average -496.67 0.0001 1 996 Constant 7489.80 0.0000 Ap-clay 135.93 0.0029 75 Others report that yield variability within a given year is controlled by soil properties that affect patterns in plant available water holding capacity or soil drainage and aeration. Mulla et al. (1992) studied spatial patterns in properties affecting winter wheat grain yield. The properties with the highest correlation to variation in yield were soil profile available water content and organic matter. Jaynes et al. (1995) also found that spatial variations in corn yield were correlated with soil properties that affected variations in soil moisture, especially in years that were excessively wet. Thus, a close relationship can be expected between soil properties that affect variations in soil moisture and crop yields. In this study, however, most of the yield variability was not explained by soil physical properties or by water availability as measured by water table depth or water content measurements over the growing season. Stepwise regression analysis was used to the parameters in Table 2.1 to the differences between corn yields obtained with chisel plowing and no-till. For 1995, 50% of the yield difference between tillage systems was related to a linear combination of VMC-tasseling, VMC-average, and the water table depth at tasseling (WT -tasseling; Table 2.8). Table 2.8. Results of stepwise regression of the difference between no-till and chisel plow corn grain yield and soil properties for 1995. Variables Coefficient P-value Constant 5276.45 0.001 1 VWC-tasseling 224.63 0.0001 VWC-average -415.51 0.0106 WT-tasseling -l674.31 0.0099 76 F rom this relationship it appears that no-till was higher yielding than chisel plowing in 1995 when season average surface water contents were lower (lower water contents in the early part of the growing season in 1995) and water at tasseling was more available (the surface water content was higher and the water table was closer to the surface). No other linear combinations of regression variables in Table 2.1 were significant in explaining the variability in corn yield in 1995 and none were significant for 1996. The general lack of significant relationships between corn yield and the parameters in Table 2.1 is not surprising given the dynamic nature of the factors that regulate crop yield. Crop simulation models were developed to capture the dynamics of crop growth and development. Chapter 3 describes the evaluation of the CERES-maize model (DSSAT version 3, 1994) to account for the spatial variability in corn yield using the data measured in this study. CONCLUSION Tillage altered soil physical properties and soil water availability primarily by destroying or reducing the spatial structure developed under no tillage. Tillage also altered the spatial pattern of corn grain yield but was different each year. While chisel plowing produced higher corn yields in 1995 for the field, no-till produced higher yields in some portions of the field in both years. Yield differences between tillage systems varied spatially, with spatial patterns differing each year. Soil properties and water availability measurements were of little value in explaining corn yields in either year using multiple regression techniques. Crop simulation models, however, might be better 77 suited to this task and should be evaluated as to their ability to predict the spatial variability in crop yield over time. A considerable effort was made to measure factors that might affect corn grain yields, but a substantial amount of spatial and temporal variability was found. The information from this study suggests that com grain yield variability can be substantial from year to year and between tillage systems. The lack of grain yield stability on the study area causes some concern for practical applications of this information on site- specific management. Several years worth of yield maps will be needed to define site- specific soil management zones in this experimental area. CHAPTER 3 Simulation of Within Field Variability of Corn Yield with CERES-Maize INTRODUCTION Simulation models are valuable tools for predicting crop yields and examining alternative agricultural management practices for crop production and their potential impacts on the environment. Simulation models are used for both scientific research purposes and to contribute to decision-making processes at experimental and governmental levels (Addiscott, 1993). Intuitively, crop models should have considerable value in evaluating site-specific management (S SM) systems and associated component precision farming practices but have found limited use (Van Uffelen et al., 1997; Han et al., 1995). Sadler and Russell (1997) suggested that limited application appears to be caused by lack of knowledge about within-field variability of soil properties needed for predicting crop yield. Traditionally, models were developed assuming soil properties to be homogenous (Han etal., 1995). Several modeling attempts have been reported using soil map delineation as a means to express spatial variability patterns, using “representative” soil profiles for each delineation (Kiniry et al., 1997; Pang et al., 1997; Gabrielle and Kengni, 1996). However, the assrunption that delineated areas on a second order soil map can serve to adequately represent soil spatial variability is flawed because considerable variation occurs within map units (Beckett and Webster, 1971). Nonetheless, since SSM 78 79 aims to maximize crop production through efficient use of managed inputs according to localized variability in soils, pests, and crop condition (Pierce and Sadler, 1997), model inputs should include details on within field variability. If soils in a field exhibit large spatial variability, many different management zones may be identified for SSM. According to Verhagen and Bouma (1997), modeling results can be very useful for evaluating effects of spatial and temporal soil variability on crop yield because they can cover a wide range of conditions that are relevant to agricultural production and environmental quality concerns. The ability of models to predict yield based on soil variability on a finer spatial scale still needs to be demonstrated in order to strengthen confidence in their usefulness for SSM (Sadler and Russel, 1997). Therefore, results of simulations must be validated with real data, obtained by field measurements on a finer scale and several growing seasons (V erhagen and Bouma, 1997). Interpolation techniques make it possible to extend results of simulations obtained at point locations to large land areas (F inke, 1993). Therefore, simulations over the range of annual weather conditions expected for a given area could be used to delineate areas displaying consistent patterns over time that can be used for SSM. Both complex mechanistic models and simple functional ones have been used to estimate crop yields. The Decision Support System for Agrotechnology Transfer (DSSAT) integrates several models, which are in the functional class, with standardized input and output (IBSNAT, 1989). The DSSAT maize model (CERES-Maize) has been tested and used in the USA and around the world with promising results. Kiniry et al. (1997) tested the model at one county in each of the nine states in the US. (Minnesota, 80 New York, Iowa, Illinois, Nebraska, Missouri, Kansas, Louisiana, and Texas) and concluded it was appropriate for predicting yield for most counties. Hodges et al. (1987) found that earlier versions of CERES-Maize accurately simulated maize grain yield in the northern U.S. Corn Belt. Researches have adapted the model in Michigan (Algozin et al., 1988), California (Pang et al., 1997), France (Gabrielle and Kengni, 1996), and China (Wu et al., 1989). While CERES-Maize model has been used successfully in many situations, like most crop simulation models, it has not been tested under conditions where soil spatial variability within a field was taken into account. A successful application of CERES- Maize to known conditions of spatial variability over time would encourage the use of crop simulation models in SSM. This study used the CERES-Maize model to evaluate the potential for proven crop simulation models to account for known soil spatial variability on simulated corn yield within a field. MATERIAL AND METHODS The DSSAT version 3 (1994) of the CERES-Maize model was applied to an extensive data set consisting of soil profile properties and corn performance data (1995 and 1996) obtained on a 30.5 m grid positioned in a 3.9 ha area located within a larger field 6 km south of Durand, Michigan as described in Chapters 1 and 2 (Figure 3.1). The model was first calibrated on one site (soil profile #15) chosen because it was nearly level, representative of the soil mapping unit, the water table not present within 1.5 m, and plant populations were similar and close to the target in both years (1995 and 1996). 81 “5‘51 Northing (m) 474760 c ‘_ _. new-t- -«e-Ir'vn- -“M‘ an. m .mm. .. 1.. _.::.\ .22 I 257200 257800 257400 257500 257600 257700 Easting (m) Figure 3.1. Location of the grid soil profile samples relative to the map units from the Order two soil survey. Refer to Table 1.1 for map units descriptions. 82 Soil profile characteristics of profile 15 as used in the model calibration are given in Table 3.1. The model was calibrated for 1995 only in order to establish appropriate crop coefficients for the corn hybrid used both years. Table 3.1. Soil characteristics of the profile 15 used in the calibration of CERES-Maize model. Horizon Clay' Silt Sand BD‘ DULll LL1 SATT OM" depth (cm) 0-25 12 24 64 1.51 0.20 0.05 0.35 1.26 25-49 19 16 64 1.70 0.24 0.1 0.33 0.7 49-75 31 22 48 1.66 0.30 0.17 0.34 0.5 75-100 22 31 46 1.69 0.28 0.15 0.33 0.4 'Clay, silt and sand = (%); 8131) = bulk density (Mg mi); *LL = lower limit of available soil water (m3 m'3); ‘DUL = drained upper limit of available soil water (m3 m'3); ISAT = saturated limit of available soil water (m3 m'3) “OM = organic matter (%). Input to the model included on-site weather data collected both years, soil data (soil profile properties), soil N balance parameters, crop management data, and genetic parameters and crop coefficients for the corn hybrid used. The minimum set of weather data (daily minimum and maximum temperatures, solar radiation, and rainfall) was collected by a minimum data weather station established at the experimental area. The soil N balance parameters and crop management data were obtained from the characteristics of the site described in Chapter 2. Plant nutrients were not a limitation factor for the present study, and pests (insects, diseases, and weed) were controlled and 83 posed no limitations to crop growth and yield. The genetic parameters: the thermal time from seed emergence to the end of the juvenile stage (Pl), the thermal time from silking to physiological maturity or black layer formation (PS), and the photoperiod sensitivity coefficient (P2) were calibrated with real time periods observed during the growing season. The maximum kernel number per plant parameter (G2) and potential kernel growth rate parameter, (G3) were set as 650 and 9.6 mg kernel" d“. The values of G2 and G3 were close to those for corn hybrids grown in the northern U.S. Corn Belt. As described in Chapter 1, other data were obtained at each grid site including water table depth measured weekly throughout the growing season using piezometers installed at each grid point to a depth of 1.5 m, volumetric soil water content in the surface 18 cm measured weekly by time domain reflectometry (TDR), and soil profile volumetric water content (soil profile storage water) monitored weekly at every other 30.5 m grid point by neutron probe access tubes, 1.5 m deep. From these measurements, soil factors needed for model input related to soil water were obtained including: drainage coefficient (SWCON), lower limit of the available soil water (LL), drained upper limit of the available soil water (DUL) or field capacity, potential extractable soil water (PLEXW), and saturated water content (SAT). These parameters were also calculated according empirical equations based on the texture of each horizon (Ritchie 1997, personal communication) and compared to the measured data set. The SCS curve number and soil albedo were determined for each grid point (Ritchie et al., 1990). The initial soil water content for simulation was set to field capacity. Once the model was calibrated for a grid point 15, it was used to simulate corn yield for all grid points for the two years (1995 and 1996), varying the profile soil data for 84 each grid point, the weather for each respective year, and using plant populations measured at harvest as model input. The performance of the CERES-Maize model was evaluated by regressing actual corn yields measured for each grid location within the experimental site (Chapter 2) with yields predicted by the model. RESULTS AND DISCUSSION Model Calibration The calibration of the CERES-Maize model on profile 15 required a few changes to the initial model parameters. The drainage rate was set to 0.5 d", based on estimates by the procedure of Ritchie and Crum (1989). Soil albedo was set to 0.13 (Ritchie et al., 1990). The SCS runoff curve number was set to 67 (Ritchie et al., 1990). The P1 genetic parameter was set a 200 degrees day, and P5 was set at 680 degrees days to produce a corn growing period consistent with the experimental site. The P2 genetic parameter was set at 0.5 h". These input values gave a harvest index (ratio of total biomass to grain yield) close to 0.5, which is expected for corn grain yield when neither nutrients nor water are limiting. Plant population for the grid point 15 was 60000 plants ha'l for 1995. The model predicted corn grain yield well for the calibration year differing by only -2 % for 1995. These differences were considered reasonable for model performance and the model was considered ready for use in the evaluation of spatial variability. 85 Simulation of Spatial Variability Corn grain yields simulated by CERES-Maize corresponded well to measured yields in 1995 (Figure 3.2-3.4). The simulated yields generally fall on the 1:1 line and a simple linear regression of simulated on measured yield accounted for 86% variability (Figure 3.2). There is a tendency of the model to underpredict at yields above 9 000 kg ha'l (Figure 3.2) and this is made clear in the comparison of frequency diagrams of measured and simulation yields in Figure 3.3. The average, standard deviation and CV for the measured and simulated data for the whole area (33 profiles) are very similar, with the difference in mean of only 27 kg ha' ‘ (Figure 3.3). The yield maps interpolated fiom measured and simulated yields are comparable (Figure 3.4a,b), with maps show similar patterns within the area, i.e., lower yields in the eastern part of the field and higher yields in the central southern part. Differences between yield maps were generally within 3: 300 kg ha'1 (Figure 3.4c). The performance of CERES-Maize in predicting the corn yield variability within the experimental area for 1995 is encouraging and is consistent with earlier evaluations of the model (Kiniry et al., 1997). Simulations for 1996 were performed using the same model coefficients and soil data sets used for 1995 and using the 1996 weather data and 1996 plant population for each grid point. The CERES-Maize model predicted very low grain yields for 1996, with an average of 1167 kg ha‘1 and a range of 469 to 1946 kg ha", while the measured yields averaged 8928 kg ha‘l and ranged from 7180 to 10 159 kg ha". Plant population for 1996 was similar to that for 1995. 86 10000 (D O O O l 8000 — 7000 — 6000 — Simulated corn grain yield (kg ha") 5000 1:1Hne 5000 l 6000 Measured corn grain yield (kg ha'1) T l 7000 8000 l 9000 10000 Figure 3.2. Plot of measured and simulated corn grain yield for 1995. 87 10 Mean = 8302 Measured 8 __ STD = 701.3 __ CV = 8.4 —l ‘1 r a — t1) 3 O’ 9 4 — LL. .1 ._ 2 _ 0 D H H Fl H 16 - — Mean = 3275 Simulated 14 - STD = 599.3 12 _. cv— 7.2 g 10 — fl 8 cr 8 _. 9 L1. _ 4 _ 2 __ O 1—1 l [—I l I—Ll l l l l 5500 6000 6500 7000 7500 8000 8500 9000 950010000 Corn Grain Yield (kg ha") Figure 3.3. Frequency distribution, mean, standard deviation (STD), and coefficient of variation (CV) for measured and simulated corn grain yield for 1995. Northing (m) Northing (m) Northing (m) 478 ' 7////2/fi1 / / / ,./ I /, V / / I \ \ {{x‘: 9000 , \ / (A . x , \ x, 257 257450 25 575 2576 2 e 0111319.) 2W1 400 .\ ' . 4780 lllll"/\\ , , §\\\/;/%l I 57 57500 257550 257600 257650 7000 Figure 3.4. Measured (A), simulated (B). and difference between measured and simulated (C) of corn grain yield (kg/ha) for 1995. 89 The model greatly under predicted grain yields because of a drought during grain fill from 1 to 19 August (Julian date 213 to 231; Figure 3.5), during which cumulative rainfall was 3 mm. Regular field observations made during this period indicated no apparent plant water stress suggesting water was not limiting to the corn during this period. Water was available to the plant from the upward flux of water from the water table as evidenced by the map of the maximum water table depth for 1996 (Figure 3.6). Even for profile 15, which was selected for model calibration because the water table was generally not within 1.5 m, water was available to the plant below 75 cm even though the upper profile water content was at the lower limit of water availability during this dry period (Figure 3.7). The problem for the CERES-Maize model is that it does not account for the presence of a water table and the upward flux of water into the root zone. Since water did not appear to be limiting in this field in 1996, the simulations were rerun with the model set to eliminate water stress during the growing season. Thus, simulated corn grain yields were determined primarily by weather conditions and the plant populations determined at harvest. Under these conditions, the model did predict the average corn yield well (8717 versus 8948 kg ha") but under predicted the variance and the frequency of yield (Figure 3.8). For 1996 under a no water stress condition, the model over predicted yield below 9000 kg ha" (above the 1:1 line in Figure 3.9) and under predicted yields above 9000 kg ha" (below the 1:1 line in Figure 3.9). The indication of under prediction at higher yields was evident in 1995 as well. 90 80 - 70 5 60 2 50 —— 40 4 30 — Rainfall (mm) 204 10- OJ 223 235 247 259 271 283 295 O) f- 0) ‘- s- N Julian date 9 103 115 127 139 151 163 75 187 F Figure 3.5. Precipitation recorded at the experimental area from April 1“ to October 31“, 1996. 92 O 5 10 15 20 25 30 35 O 1 l 20 - 40 A g, 60 _. E. o. g 80 _ —- LL15 —I— 8/9/96 + 8/19/96 120 — + 8/28/96 Figure 3.7. Volumetric soil water content (%) at drainage upper limit and lower limit for the grid point 15 and volumetric soil water content in different dates 93 7 6 Mean = 8948 Measured STD = 776.0 F 5 _ CV= 8.7 _ >4 8 4 _ (D 3 g, 3 _ fl _ LL 2 — 1 —- "—1 — 7 "l 1“ WI Ti ll 0 Simulate 10 ‘ Mean = 8717 STD = 365.3 —- 3 2 CV=4.2 6 __ C 8 6 _ a. 92 LL 4 1 fl 2 " ll 0 l lfl l I j j 7000 7500 8000 3500 9000 9500 10000 10500 Corn grain yield (kg ha") Figure 3.8. Frequency distribution, mean, standard deviation (STD), and coefficient of variation (CV) for measured and simulated corn grain yield for 1996. 94 .7" 10000 2 m .c 5'3 2 .1; ‘ _, e g 9000 . ..' .5.’e ‘2’ 0 $ e' e e s. ‘ . C 8 '0 - ' % 8000 fl . 1.1 11116 S E e e '0') 7000 I I I 7000 8000 9000 1 0000 Measured corn grain yield (kg ha") Figure 3.9. Plot of measured and simulated corn grain yield for 1996. 95 "ragrggirggr' ram Egg; llllllllllllIlm ‘ W585 l" l téfi§§§l‘y4£l arm Veeeeeesxess fill" lllllllllllll 42‘ '4' .z: iii 1’ in Northing (m) ,agiiifféefiieesa . ,8 H" ..r: , . mi: “‘5 "34'3“” A WearilifiieaialullllIlllllllllllllllllllllllllIIIIIIIIllllll'i\\\\\\\\\\s .. . - 500 257450 257500 257550 NH” 000 Easting (m) llfill"""1"“lllllllllllll”:~ i 0’01” . "ifigbff 1::i,E,’§,§§: 5‘: 1 1 ' 51-11 9153' 11} gig; 131311.595” :5: . . fig???wiggfiagjfiéggg’yfaggfim.i 4 fig; . I: 1'.: ... I: i}? ' 111513 111: ‘li'gsiigsir, r g r," r r :8. 5. , n : a: .. if” '5‘? 1' ‘ * tiff; “59333485 . ii litigiii ;. ~§ ‘r. Northing (m) 4 i 3% f,“ grurr .. .1 1E111111111111111: X r 257400 257450 257500 257550 257600 6000 Easting (m) 257500 257550 257600 257650 A // . Northing (m) Easting (m) -300 Figure 3.10. Measured (A), simulated (B), and difference between measured and simulated (C) of corn grain yield (kg/ha) for 1996. 96 For 1996, the results indicate that variation in soil properties in combination with differential effects of the water table affected the spatial distribution of corn yield not accounted for in the model. There is little correlation between the yield map interpolated from the measured data and that interpolated from the simulated yields, although yield differences were generally within 600 kg ha’1 (Figure 3.10). CONCLUSION For a single year, 1995, the CERES-Maize model simulated corn yields and yield variability well, accounting for a large portion of the variability in yield and producing a yield map very similar to the measured values. A drought during 1996 showed the importance of water table in supplying water to corn within this field. The CERES-Maize does not account for water table and will need modification for soils where water table contributions are significant to crop yield. These results show the potential for crop models to simulate the spatial and temporal variability in yields and clearly demonstrate the importance of the presence of a water table in understanding variability in crop yield and in the delineation of management zones within fields for site-specific management. LIST OF REFERENCES REFERENCES Addiscott, T.M. 1993. Simulation modeling and soil behavior. Geoderma 60: 15-41. Addiscott, T.M. and R.J. Wagenet. 1985. Concepts of solute leaching in soils: a review of modelling approaches. J. Soil Sci. 36:411-424. Algozin, K.A., V.F. Bralts and J .T. Ritchie. 1988. Irrigation strategy selection based on crop yield, water, and energy use relationships: A Michigan example. J. Soil Water Conserv. 43 :428-43 1. Azooz, R.H., M.A. Arshad, and A.J. Franzluebbers. 1996. Pore size distribution and hydraulic conductivity affected by tillage in northwestern Canada. Soil Sci. Soc. Am. J. 60:1197-1201. Beckett, P.H.T. and R. Webster. 1971. Soil variability: a review. Soil and Fertil. 34:1-15. Blake, GR. and K.H. Hartge. 1986. Bulk density. p.363-375. In A. Klute (ed.) Methods of soil analysis. Part 1. 2nd ed. Agron. Monogr. 9. ASA, Madison, WI. Blevins, R.L., M.S. Smith, and W.W. Frye. 1983. Changes in soil properties after 10 years of no-tillage and conventional tilled corn. Soil Tillage Res. 3:135-146. Bouma, J. and PA. Finke. 1993. Origin and Nature of Soil Resource Variability. p.3-13. In P.C. Robert et al. (ed) Procedures of Soil Specific Crop Management: A Workshop on Research and Development Issues. SSSA Spec. Publ. Soil Sci. Soc. Am., Madison, WI. Boyer, D.G., R.J. Wright, C.M. Feldhake, and D. P. Bligh. 1996. Soil spatial variability relationships in a steeply sloping acid soil environment. Soil Sci.. 161:278-287. Cambardella, C.A., T.B. Moorrnan, J .M. Novak, T.B. Parkin, D.L. Karlen, R.F. Turco, and A.E. Konopka. 1994. Field-scale variability of soil properties in central Iowa soils. Soil Sci. Soc. Am. J. 58:1501-1511. Caron, J. C.R. Espindola, and DA. Angers. 1996. Soil structural stability during rapid wetting: Influence of land use on some aggregate properties. Soil Sci. Soc. Am. J. 60:901-908. Carter, MR. 1995. Spatial variability of soil porosity under reduced tillage in a Humo- Ferric Podzol. Can. J. Soil Sci. 75:149-152. Chien, Y., D. Lee, H. Guo, and K. Houng. 1997. Geostatistical analysis of soil properties of mid-west Taiwan soils. Soil Sci. 162:291-298. 97 98 Chung, C.K., S. Chong, and EC. Varsa. 1995. Sampling strategies for fertility on a stoy silt loam soil. Commun. Soil Sci. Plant Anal. 26:741-763. Ciha, A.J. 1984. Slope position and grain yield of soft white winter wheat. Agron. J. 76: 193-196. Elliott, E.T., I.C. Burke, C.A. Monz, S.D. Frey, K.H. Paustian, H.P.Collins, E.A. Paul, C.V. Cole, R.L. Blevins, W.W. Frye, D.J. Lyon, A.D. Halvorson, D.R Huggins, R.F. Turco, and M.V. Hickman. 1994. Terrestrial carbon pools: Preliminary data fi'om the Corn Belt and Great Plains regions. p. 179-191. In Doran, et al. (eds.) Defining soil quality for a sustainable environment. SSSA Spec. Publ. No 35. Soil Sci Soc. Am., Madison, WI. Entz, T. and C. Chang. 1991. Evaluation of soil sampling schemes for geostatistical analyses: A case study for soil bulk density. Can. J. Soil Sci. 71 :165-176. Everett, M.W. and F .J . Pierce. 1996. Variability of corn yield and soil profile nitrates in relation to site-specific management. p. 43-53. In P.C. Robert et al. (ed.) Precision Agriculture. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI. Finke, PA. 1993. Field scale variability of soil structure and its impact on crop growth and nitrate leaching in the analysis of fertilizing scenarios. Geoderma 60:89-107. Finke, P.A., J. Bouma, and A. Stein. 1992. Measuring field variability of disturbed soils for simulation purpose. Soil Sci. Soc. Am. J. 56:187-192. Gabrielle, B. and L. Kengni. 1996. Analysis and field-evaluation of the CERES models’soil components: nitrogen transfer and transformations. Soil Sci. Soc. Am. J. 60: 142-149. Gaunt, J.L., J. Riley, A. Stein, and F .W.T. Penning de Vries. 1997 . Requirements for effective modelling strategies. Agric. Systems 54:153-168. Gee, GE. and J .W Bauder. 1986. Particle-size Analysis. p.383-410. In A. Klute (ed.) Methods of soil analysis. Part 1. 2nd ed. Agron. Monogr. 9. ASA, Madison, WI. Griffith, D.R., J .V. Mannering, and J .E. Box. 1986. Soil and moisture management eith reduced tillage, p. 19-55. In M.A. Sprague and GB. Triplett (ed.) No tillage and surface-tillage agriculture. Willey, New York. Guillermo, A.S., H.E. Echeverria, and EM. Casanovas. 1997. Crop-pasture rotation for sustaining the quality and productivity of a Typic Argiudoll. Soil Sci. Soc. Am. J. 61:1466-1472. Gupta, V.V.S.R., P.R. Grace, and M.M. Roper. 1994. Carbon and nitrogen mineralization as influenced by long-term soil and crop residue management systems in Australia. p. 99 193-200. In J .W. Doran et al (eds) Defining soil quality for a sustainable environment. SSSA Spec. Publ. No 35. Soil Sci Soc. Am., Madison, WI. Han, S., R.G. Evans, T. Hodges, and S.L. Rawlins. 1995. Linking a geographic information system with a potato simulation model for site-specific management. J. Environ. Qual. 24:772-777. Han, 8., SM. Schneider, R.G. Evans, and S.L. Rawlins. 1996. Spatial variability of soil properties on two center pivot irrigated fields. p. 97-106. In P.C. Robert et al. (ed.) Precision Agriculture. ASA Misc. Publ., ASA, CSSA, and SSSA, Madison, WI. Hanks, J. and J .T. Ritchie. 1991. Modeling plant and soil systems. Agron. Monogr.31. ASA, CSSA, and SSSA, Madison, WI. Hanna, A.Y., P.W. Harlan, and D.T. Lewis. 1982. Soil available water as influenced by landscape position and aspect. Agron. J. 74:999-1004. Hergert, G. W., W.L. Pan, D.R. Huggins, J .H. Grove, and TR. Peck. 1997. Adequacy of current fertilizer recommendations for site-specific management. p. 283-300. In F.J. Pierce and E]. Sadler. (ed.) The state of site-specific management for agriculture. ASA Misc. Publ., ASA, CSSA, and SSSA, Madison, WI. Hill, R.L and M. Meza-Montalvo. 1990. Long-term well traffic effects on soil physical properties under different tillage systems. Soil Sci. Soc. Am. J. 54:865-870. Hodges, T., D. Botrrer, C. M. Sakamoto, and J. Hays Hang. 1987. Using the CERES- Maize model to estimate production for the US. Corn-Belt. Agric. for Metereol. 40:293-303. Hubbard, R.K., W.L. Hargrove, R.R. Lowrance, R.G. Williams, and B.G. Mullinix. 1994. Physical properties of a clayey coastal plain soil as affected by tillage. J. Soil and Water Conserv. 49:276-283. IBSNAT. 1989. Decision Support System for Agrotechnology Transfer (DSSAT)-User’s Guide. IBSNAT Project, Dept. Agronomy and Soil Sci., University of Hawaii, Honolulu. Issaks, EB. and RM. Srivastava. 1989. An introduction to applied geostatistics. Oxford Univ. Press, Oxford. Jaynes, D.B., T.S. Colvin, and J. Ambuel. 1995. Yield mapping by eletromagnetic induction. p. 383-394. In P.C. Robert et al. (ed.) Site-specific management for agricultural systems. ASA, CSSA, and SSSA, Madison, WI. 100 Jones J .W. and J .T. Ritchie. 1991. Crop growth models. In G.J. Hoffman, T.A. Howell, and K.H. Soloman (eds) Management of farm irrigation systems. ASAE, St Joseph, MI. Journel, AG. and C.J. Huijbregts. 1978. Mining geostatistics. Academic Press, London. Kemper, W.D. and RC. Rosenau. 1986. Aggregate stability and size distribution. p. 425- 442. In A. Klute (ed.) Methods of soil analysis. Part 1. 2nd ed. Agron. Monogr. 9. ASA, Madison, WI. Khakural, B.R., G.D. Lemme, T. E. Schumacher, and M.J. Lindstron. 1992. Effects of tillage systems and landscape on soil. Soil Till. Res. 25:43-52. Khakural, B.R., P.C. Robert, and DJ. Mulla. 1996. Relating corn/soybean yield to variability in soil and landscape characteristics. p. 117-128. In P.C. Robert et al. (ed.) Precision Agriculture. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI. Kiniry, J .R., J .R.Williams, R.L. Vanderlip, J .D. Atwood, D.C. Reicosky, J. Mulliken, W.J. Cox, H.J. Mascagni Jr., S.E. Hollinger, and W.J. Wiebold. 1997. Evaluation of two maize models for nine U.S. locations. Agron. J. 89:421-426. Kitur, B.K., KR Olson, J .C. Siemens, and SR. Phillips. 1993. Tillage effects on selected physical properties of Grantsburg silt loam. Commun. Soil Sci. Plant Anal. 24:1509- 1527. Klute, A. 1986. Water retention: Laboratory methods. p. 635-662. In A. Klute (ed.) Methods of soil analysis. Part 1. 2"d ed. Agron. Monogr. 9. ASA, Madison, WI. Lamb, J .A., R.H. Dowdy, J .L. Anderson, and G.W. Rehm. 1997. Spatial and temporal stability of corn grain yield. J. Prod. Agric. 10:410-414. Larson, W.B. and P.C. Robert. 1991. Farming by soil. p. 103-112. In R. Lal and F.J. Pierce (ed.) Soil management for sustainability. SWCS, Ankeny, Iowa. Lindstrom, M.J., W.B. Voorhees, and TE. Schumacher. 1995. Soil properties across a landscape continuum as affected by planting wheel traffic. p. 351-363. In P.C. Robert et al. (ed.) Site-specific management for agricultural systems. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI. Mallants, D., B.P. Mohanty, D. Jacques, and J. F eyen. 1996. Spatial variability of hydraulic properties in a multi-layered soil profile. Soil Sci. 161 :167-181. Mausbach, M.J., D.J. Lytle, and L.D. Spivey. 1993. Application of soil survey information to soil specific farming. p. 57-68. In P.C. Robert et al. (ed) Proc. of soil specific crop management: A workshop on research and development issues. SSSA Spec. Publ. Soil Sci. Soc. Am., Madison, WI. 101 McGraw, T. and R. Hemb. 1995. Fertility variability in the Minnesota River Valley Watershed in 1993 as determined from grid testing results on 52,00 acres in commercial fields. p. 167-174. In P.C. Robert et al. (ed.) Site-specific management for agricultural systems. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI. Moore, I.D., P.E. Gessler, G.A. Nielsen, and GA Peterson. 1993. Terrain analysis for soil specific crop management. p. 27-52. In P.C. Robert et al. (ed) Proc. of Soil specific crop management: A Workshop on research and development issues. SSSA Spec. Publ. Soil Sci. Soc. Am., Madison, WI. Mulla, D.J., A.U. Bhatti, M.W. Hammond, and J .A. Benson. 1992. A comparison of winter wheat yield and quality under uniform versus spatially variable fertilizer management. Agric. Ecosys. Environ. 38:301-31 1. Mulla, D.J., and J .S. Schepers. 1997. Key processes and properties for site-specific soil and crop management. p. 1-18. In F.J. Pierce and E.J. Sadler. (ed.) The state of site- specific management for agriculture. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI. Nix H.A., 1983. Minimum data sets for agrotechnology transfer. Proceedings of the International Symposium on Minimum Data Sets for Agrotechnology Transfer. 21-23 March. ICRISAT AP 502324. India. Pang, X.P., J. Letey, and L. Wu. 1997. Yield and nitrogen uptake prediction by CERES- Maize model under semiarid conditions. Soil Sci. Soc. Am. J. 61 :254-256. Pannatier, Y. 1996. Variowin: software for spatial data analysis in 2D. Springer-Verlag, New York, Inc., Edwards Brothers, Inc., Ann Arbor, MI. Passioura J .B., 1996. Simulation Models: Science, snake oil, education and engineering? Agron. J. 88:690-695. Pierce, F.J. and DD. Warncke. 1994. Site-specific management: An overview. p.3-6. In Site-specific management in agriculture. Proc. Workshop. Kellogg Center, MSU, East Lansing, MI. March, 7. 1994. Pierce, F.J., D.D. Warncke, and M.W. Everett. 1995. Yield and nutrient variability in glacial soils of Michigan. p. 133-154. In P.C. Robert et al. (ed.) Site-specific management for agricultural systems. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI. Pierce, F.J. and E.J. Sadler (ed.). 1997. The state of site-specific management for agriculture. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI. 102 Pierce, F .J ., M. C. F ortin, and M.J. Staton. 1994. Periodic plowing effects on soil properties in a no-till farming system. Soil Sci. Soc. Am. J. 58: 1782-1787. Poier, K.R. and J. Richter. 1992. Spatial distribution of earthworms and soil properties in an arable loess soil. Soil Biol. Biochem. 24:1601-1608. Reinert, DJ. 1990. Soil structural form and stability induced by tillage in a Typic Hapludalf. Ph.D. diss. Michigan State University, East Lansing, MI. Richardson, CW. and K.W. King. 1995. Erosion and nutrient losses from zero tillage on a clay soil. J. Agric. Eng. Res. 61:81-86. Ritchie, J .T., D.C. Godwin, and S. Otter-Nacke. 1985. CERES-Wheat: A simulation model of wheat growth and development. Texas A&M University Press, College Station, TX. Ritchie, J .T., D.C. Goldwim, and U. Singh. 1990. Soil and weather inputs for the IBSNAT crop models. p. 31-45. In Procedures of the IBSNAT symposium: Decision support system for agrotechnology transfer. Part 1; Symposium Proceedings. Las Vegas, NV. 16-18 Oct. 1989. Dept. of Agronomy and Soil Science, College of Tropical Agriculture and Human Resources, University of Hawaii, Honolulu, HI. Ritchie, J .T., and J. Crum. 1989. Converting soil survey characterization data into IBSNAT crop model input. p. 155-167. In J. Bouma and A.K. Bregt (ed.) Land qualities in space and time. PUDOC, Wageningen, The Netherlands. Ritchie, J.T., U. Singh, D.C. Godwin and L. Hunt. 1989. A user’s guide to CERES- Maize-V.2.10. Muscle Shoals, AL. International Fertilizer Development Center. Robert, P. 1993. Characterization of soil conditions at the field level for soil specific management. Geoderma 60:57-72. Robert, P. C., R.H. Rust, and W.B. Larson. 1993. Proceedings of soil specific crop management: A workshop on research and development issues. SSSA Spec. Pibl. Soil Sci. Soc. Am., Madison, WI. Sadler, E.J. and G. Russell. 1997. Modeling crop yield for site-specific management. p. 69-79. In F.J. Pierce and E.J. Sadler. (ed.) The state of site-specific management for agriculture. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI. SAS Institute 1990. SAS/STAT user’s guide. Version 6. SAS Inst, Cary, NC. Seta, A.K., R.L. Blevins, W.W. Frye, and B.J. Barfield. 1993. Reducing soil erosion and agricultural chemical losses with conservation tillage. J. Environ. Qual. 22:661-665. 103 Spomer, R.G., and R.F. Piest. 1982. Soil productivity and erosion of Iowa loess soils. Trans. ASAE 25: 1295-1299. Staricka, J .A., R.R. Almaras, and W.W. Nelson. 1991. Spatial variation of crop residue incorporated by tillage. Soil Sci. Soc. Am. J. 55:1668-1674. Stone, J.R., J .W. Gilliam, D.K. Cassel, RB. Daniels, L.A. Nelson, and H.J. Kleiss. 1985. Effects of erosion and landscape position on the productivity of Piedmont soils. Soil Sci. Soc. Am. J. 49:987-991. Sudduth, K.A., J .W. Hummel, and SJ. Birrel. Sensors for site-specific management. p. 183-210. In F.J. Pierce and E.J. Sadler. (ed.) The state of site-specific management for agriculture. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI. Swayer, J .E. 1994. Concepts of variable rate technology with considerations for fertilizer application. J. Prod. Agric. 7:195-201. Threlkeld, G.W. and J .E. Feenstra. 1974. Soil Survey of Shiawassee County, Michigan. US. Gov. Print Office, Washington, DC. Trangmar, B.B., R.S. Yost, and G. Uehara. 1985. Application of geostatistics to spatial studies of soil properties. Advances in Agronomy. 38:45-94. Van Uffelen, C.G.R., J. Verhagen, and J. Bouma. 1997. Comparison of simulated crop yield patterns for site-specific management. Agric. Syst. 54:207-222. Verhagen, J. and J. Bouma. 1997. Modeling soil variability. p. 55-67. In F.J. Pierce and E.J. Sadler. (ed.) The state of site-specific management for agriculture. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI. Voorhees, W.B., R.R. Allrnaras, and M.J. Lindstron. 1993. Tillage considerations in Managing Soil Variability. p. 95-111. In P.C. Robert et al. (ed) Procedures. of Soil Specific Crop Management: A Workshop on Research and Development Issues. SSSA Spec. Publ. Soil Sci. Soc. Am., Madison, WI. Waddell, J .T. and RR. Wei]. 1996. Water distribution in soil under ridge-till and no-till corn. Sci. Soc. Am. J. 60:230-237. Webster, R. Quantitative spatial analysis of soil in the field. 1985. Advances in Soil Sci. 3: 1-70. Webster, R. and MA. Oliver. 1990. Statistical methods in soil and land resource survey. Oxford Univ. Press, New York. Wollenhaupt, D. J. Mulla, and CA. Gotway Crawford. 1997. Soil sampling and interpolation techniques for mapping spatial variability of soil properties. p. 19-53. In F.J. 104 Pierce and E.J. Sadler. (ed.) The state of site-specific management for agriculture. ASA Misc. Publ. ASA, CSSA, and SSSA, Madison, WI. Yule, I.J., P.J. Cain, E.J. Evans, and C. Venus. 1996. A spatial inventory approach to farm planning. Computers and Electronics in Agriculture. 14:151-161. Wu, Y., CM. Sakamoto, and D.M. Botner. 1989. On the application of the CERES- Maize model to the North China Plain. Agric. for Meteorol. 49:9-22.