SSSAJ Grow Your Career with SSSA
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 23 May 2006
Published in Soil Sci Soc Am J 70:1210-1221 (2006)
DOI: 10.2136/sssaj2005.0039
© 2006 Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via ISI Web of Science (3)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Kay, B. D.
Right arrow Articles by Dharmakeerthi, R. S.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Kay, B. D.
Right arrow Articles by Dharmakeerthi, R. S.
Agricola
Right arrow Articles by Kay, B. D.
Right arrow Articles by Dharmakeerthi, R. S.
Related Collections
Right arrow Nutrient Management
Right arrow Plant Analysis

Nutrient Management & Soil & Plant Analysis

Integrating Soil and Weather Data to Describe Variability in Plant Available Nitrogen

B. D. Kaya,*, A. A. Mahboubib, E. G. Beauchampa and R. S. Dharmakeerthic

a Dep. of Land Resource Science, Univ. of Guelph, Guelph, ON, N1G 2W1 Canada
b Dep. of Soil Science, Univ. of Bu–Ali Sina, Hamadan, Iran
c Dep. of Soils and Plant Nutrition, Rubber Research Institute of Sri Lanka, Dartonfield, Agalawatta, Sri Lanka

* Corresponding author (bkay{at}lrs.uoguelph.ca)


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Although there are economic and environmental reasons to manage fertilizer-nitrogen (N) more effectively in variable landscapes, the impact of weather and its interaction with soil properties/landscape attributes or management practices has received little attention. The objectives of this study were to assess the magnitude of temporal and spatial variability in soil and plant N in a variable landscape under different management practices and to assess the dependence of their temporal variability on readily available weather variables such as air temperature and rainfall. The experiment was conducted from 1997 to 2003 on a simple slope under three maize (Zea mays L.) based cropping systems. Soil and shoot N were measured through the growing season and the sum used as a measure of plant available N (PAN). Values of PAN varied with year, treatment, landscape position, and year x treatment and year x treatment x position interaction terms. The effects were quantified for each management treatment using multiple regression analyses to relate PAN to soil organic carbon (OC), cumulative degree days (CDD), and cumulative rainfall (CRF) in different periods within the growing season. Plant Available Nitrogen was most strongly influenced by rainfall early in the growing season and exhibited a nonlinear response to OC and CRF. The regression model predicted spatial patterns that were generally stable when applied to historical weather data; PAN increased with OC in 12 of the 15 yr. The analyses illustrate the feasibility of combining soils and weather data to predict N dynamics in variable landscapes.

Abbreviations: CDD, cumulative degree days • CRF, cumulative rainfall • CV, coefficient of variation • DOY, day of year • N, nitrogen • NLWR, non-limiting water range • NRMSE, normalized root mean square error • OC, organic carbon • PAN, plant available N • RMSE, root mean square error • SMN, soil mineral N • Till, tillage treatment • VCE, variance component estimate • WFPS, water-filled pore space


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
FARMERS, advisory personnel and researchers recognize the need to apply fertilizer-N at rates that satisfy both economic and environmental objectives. However, the means to determine the amount of fertilizer to achieve these objectives, particularly on variable landscapes, is less evident. Considerable effort in the past decade has been directed to enhancing our understanding of the role of soil properties and/or landscape attributes on soil N dynamics and crop response to N to manage fertilizer-N according to soil or landscape characteristics (i.e., site-specific or precision management of N). However, soil N dynamics and crop response to N also vary with the weather. Less attention has been paid to the impact of weather on these processes, although its effects may be as large, or larger than those due to soil properties or landscape attributes (Lamb et al., 1997; Eghball and Varvel, 1997; Sogbedji et al., 2001).

Weather may influence the rate of mineralization of organic N, the extent of denitrification or leaching, crop uptake and yield response to fertilizer-N, and the amount of residual N left in the soil at the end of the growing season. The impact of weather on N dynamics would be expected to vary through the season. Sogbedji et al. (2001) attributed the yearly variation in yield response of maize to fertilizer-N on three different drainage classes to variation in early season precipitation. Although the timing of the occurrence of different weather effects would also be expected to vary with agroecological zones, this possibility has not been examined. The impact of weather on N dynamics may also vary with soil properties or landscape attributes. Timlin et al. (1998) found that the intra-annual differences in weather had the greatest effect on grain yield where the magnitude of the curvature in the landscape was large. Jaynes et al. (2003) were able to group areas within a field in which yields behaved similarly among years (i.e., yields were stable or unstable, high or low) and each area was found to represent characteristic landscape attributes.

The impact of weather on different components of the N cycle in a variable landscape will be problematic for site-specific management of N if the spatial patterns in these components are not predictable within and between seasons. For instance, although soil NO3 has been proposed as an index of the availability of N in the humid areas of North America (Meisinger et al., 1992), Cahn et al. (1994) found that the spatial patterns of NO3 changed during the spring. The accuracy of estimates of fertilizer requirements based on N availability indices may be further compromised by weather conditions that may alter not only soil N dynamics but also crop response to fertilizer-N after the index is measured. These observations may account for the poor correlation between the early season spatial patterns in soil NO3 and those of yield or yield response to N (Eghball et al., 2003; Mamo et al., 2003; O'Halloran et al., 2004) as well as a lack of stability in yield patterns from year-to-year (Lamb et al., 1997; Jaynes et al., 2003).

Management practices such as tillage and the use of leguminous crops can have an important impact on N dynamics. However, little of the management-related research has been done on variable landscapes and we have found no research that considered the additional interaction with weather. The interaction between management, soil properties, and weather creates an additional layer of complexity in predicting N dynamics.

While weather cannot be accurately predicted, it is conceivable that two types of weather data may be used to improve estimates of fertilizer N requirements: real-time weather data collected up to the latest time of fertilizer application and a probabilistic description of weather following fertilizer application. In both cases the weather data would have to be used in regression or deterministic models in conjunction with soil or landscape data if the impact of weather varies with soil properties or landscape attributes. Weather data collected up to the time of the latest fertilizer application may be used in models to estimate soil mineral N (SMN). In contrast, statistics on weather through the remainder of the growing season may be used along with estimates of SMN (or indices of plant available N determined earlier in the season) in deterministic models that incorporate weather-driven changes in soil N dynamics and crop response (Campbell et al., 1995). Output of such models could include probability of yield response for different fertilizer N rates at the measured or predicted level of SMN. However, their use would only be practical at the farm level if estimated or default values were provided for a large number of the input variables since these are not routinely measured (Smith et al., 1997). Irrespective of the type of weather data, the paucity of long-term studies on the effects of weather, either before or after fertilizer application, on N dynamics in variable landscapes and different management practices means that there are few data sets available to validate the models even if all the input data were available.

We have previously described the spatial and temporal variability in plant available N (Dharmakeerthi et al., 2004, 2005) and N uptake by (Dharmakeerthi et al., 2006) on a variable landscape in the cool humid environment of southern Ontario under different maize-based management practices. The spatial variability in N availability and N uptake across the landscape was found to be mainly governed by the spatial variation in OC content and the effects of management practices such as tillage and legume incorporation on N availability and N uptake did not vary across the landscape (Dharmakeerthi et al., 2005, 2006). However, large variations between years were noted. This paper focuses on the contribution of weather to variation in plant available N and its interaction with soil properties and management practices. The objectives of this study were: (1) to assess the magnitude of temporal and spatial variability in SMN and plant N in a variable landscape under different management practices, (2) to determine the extent of interaction between temporal and spatial variability in these parameters and (3) to assess the dependence of the temporal variability on readily measurable weather variables such as air temperature and rainfall.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Field Site
Details on the experimental design, treatments, sampling, and analysis have been described elsewhere (Dharmakeerthi et al., 2004, 2005, 2006), and therefore in this paper we concentrate only on the main details. The study was established on a simple slope at the Elora Research Station, (43° 39' N and 80° 35' W) which is about 23 km northwest of Guelph, Ontario. The soil was a Typic Hapludalf according to the USDA/WRB98 classification systems with an average slope of 6%. Texture of the A horizon varied from silt loam to loam. The climate at the site is characteristic of modified continental with mean daily temperatures in January and July of –6.0 and 19.4°C, respectively. Mean annual precipitation is 943 mm. Enhanced erosion and deposition following cultivation have altered the spatial variation in the depth of the A horizon that was created by pedogenic processes before the introduction of agriculture. Measurements of 137Cs suggest that as much as 15 to 35 cm of colluvium have accumulated on the foot- and toeslope positions since 1954 (VandenBygaart, 2001). In spite of extensive erosion, the texture of the A horizon was relatively uniform with clay contents that varied from between 85 and 189 g kg–1. The organic carbon (OC) content however was more variable ranging from 7.7 to 31.2 g kg–1 at a depth of 0 to 0.3 m. The pH (in 0.01 M CaCl2) was neutral to slightly alkaline.

The study was conducted from 1996 to 2003. In the preceding 3 yr (1993–1995 inclusive) the site had been planted to corn with an unrecorded amount of fertilizer-N and solid beef manure applied annually. Before that time the site had been used for a decade or more for the production of livestock feed (hay, cereals, and maize) in a variable rotation.

Experimental Design and Treatment
A 2-yr rotation was used with maize following barley (Hordeum vulgare L.). The barley was fertilized with 60 kg N ha–1 from 1996 to 1999 inclusive and this rate reduced to 0 kg N ha–1 in 2000 to 2003 inclusive. The grain was harvested but yields of the barley generally not recorded. Three management systems were used: maize following barley using no-till (barley-NT), maize following barley using conventional tillage involving spring plowing (full inversion) and secondary tillage (barley-CT), and maize following barley that had been under seeded with red clover (Trifolium pretense L.) (barley + red clover-CT). The red clover in the latter treatment was treated as a green manure crop and the barley and red clover residues plowed down in the spring. The maize hybrid that was used varied through the study: Pioneer 3902 was planted in 1997, Northrup King N17– C5 in 1998–2001, Dekalb C35–50 in 2002 and Northrup King N17-C5 in 2003. These hybrids had crop heat unit ratings varying from 2600 to 2650.

The experiment was conducted in two areas next to each other on the same landscape in alternate years to accommodate the 2-yr rotation. Thus, maize was grown in the 1997, 1999, 2001, and 2003 seasons in one area, whereas in 1998, 2000, and 2002 seasons it was grown in the adjacent area as described in Dharmakeerthi et al. (2005). Cropping treatments were re-randomized within each replication in each area after each cycle of the rotation. All measurements were made in the maize phase of the rotation. At the time of planting, 200 kg ha–1 of 0–20–20 (oxide basis) was banded at planting to meet the P and K requirements. Maize was planted down the slope and experimental plots (9 by 8 m) were established in five positions (summit, shoulder, backslope, footslope, and toeslope). The planting density was approximately 60 000 plants ha–1 with a row spacing of 0.75 m. The plots were split lengthwise into subplots (4.5 by 8 m) with one-half receiving no N fertilizer and one-half receiving 140 kg of N ha–1 applied as a broadcast treatment around the fifth visible leaf tip stage (V2 stage). The N treatments were referred to as 0N and +N treatments, respectively. Since the cropping treatments were re-randomized within a replication at the end of each rotational cycle, the 0N treatments were not consistently in the same plots throughout the study. Altogether there were three crop management systems (main plot), five landscape positions (subplot), and two fertilizer N levels (sub-subplot). The experimental design was a modified split-split plot design with three replicates.

Measurements
The weather was characterized using CDD and CRF, with data obtained from a weather station situated about 1 km away from the experimental site. The CDD was based on mean daily air temperature (°C) accumulated from the early spring after three successive days in which the mean daily air temperature was above 0°C. Values of CDD were assumed to provide an indication of seasonal differences in soil temperature. Dharmakeerthi et al. (2005) found that the CDD, based on average daily air temperature, was well correlated to equivalent values based on soil temperature at the 5-cm depth (R2 = 0.999 and P < 0.001) measured under a grass cover. Consequently they used CDD as "thermal time" to describe the accumulation of plant available N through the growing season. Values of CDD were also highly correlated with cumulative crop heat units (R2 > 0.99 and P < 0.001 for different years). Crop heat units are based on growth-limiting minimum daytime and nighttime temperatures. However microbial activity in soils begins as soil temperature increases and approaches 0°C (Dorland and Beauchamp, 1991; Kätterer et al., 1998) and therefore a lower minimum air temperature than that used for crops was considered appropriate.

The CRF was based on average daily rainfall (mm) and was calculated from the same starting date as CDD. The distribution of rainfall through the growing season was characterized by relating the CRF to CDD and determining the cumulative rainfall through the periods that broadly approximated periods of hydrologic or phenological significance. The periods that were selected were 200 to 700, 700 to 1350, 1350 to 2000, and 2000 to 2600 CDD. The period 200 to 700 CDD extended from infiltration and drainage of snowmelt to about the time of preside-dress N soil test (the third sampling). The period 700 to 1350 extended through the period of rapid vegetative growth of the maize crop. The period 1350 to 2000 bracketed the reproductive stage of maize (sixth to seventh sampling). The period 2000 to 2600 extended to physiological maturity (10th and 11th sampling).

Soil samples were collected from 0- to 0.30-m depth to determine SMN content on the 0N plots. Sampling was conducted every second week throughout the growing season from early May to late September or early October. There were a total of 10 to 11 sampling dates in each year. Sampling was done in the center of the third interrow of the experimental plots. A minimum of 10 cores were taken per plot at each sampling date using a standard 25-mm diam. soil probe and composited. Soil samples were frozen until SMN extraction occurred. After thawing, the composite sample was mixed and a subsample used to determine gravimetric water content. The remaining sample was sieved through a 4-mm sieve and a subsample of 5.0 g used for extraction of NH4–N and NO3–N using 2 M KCl. Samples collected on the second sampling event (around late May) were used for the determination of OC content.

Maize plants (aboveground biomass) were harvested every second week at the time of soil sampling, beginning about 4 wk after planting. An area equivalent to 1 m by 2 rows was harvested for dry matter determinations. These samples were dried (70°C), weighed, subsampled, and ground to pass through a 1-mm sieve before determining the N content. The total amount of plant available N (PAN) was approximated at each sampling time by summing the SMN and N in the plant. However, it is recognized that this should be considered an index of available N since it did not include SMN below 0.3 m or root N (Beauchamp et al., 2004). At the end of the growing season, plots were hand-harvested from an area equivalent to 5 m by 2 rows, samples dried at 70°C and yields (15.5% moisture) calculated. Additional details on soil and plant analyses are given in Dharmakeerthi et al. (2005).

Variance component estimates (van Es et al., 1999) were employed to assess the relative magnitude of the temporal and spatial variability in PAN (Objective 1) and the extent of interaction between temporal and spatial variability (Objective 2). A mixed model (including fixed and random effects) with a split plot design was used with random effects for rep nested within year. The year, treatment and year x treatment interaction terms were tested with rep (year x treatment), (Error A). The remaining terms were tested with the model error (Error B). Variance component estimates (VCEs) of the main effects (year, management treatment, and landscape position) and their interaction terms were determined at four critical stages during the growing season and the magnitude of the normalized variance ([VCE]1/2 x 100/grand mean) of each term used to assess its relative contribution to the variation in PAN at the different stages in the growing season. Data for odd years (1997, 1999, 2001, and 2003) were selected for analyses so that the year effect did not include effects arising from shifting the location of the plots within the field in alternate years. Analyses were not completed for the even years because of the small number (2 yr). The four critical stages selected were: just before planting, just before the onset of rapid plant growth, at silking, and at maturity. Measurement of PAN just before planting (Sampling #1) and just before the onset of rapid growth (Sampling #3) include only SMN and correspond to values at the preplant N test (PPNT) and preside-dress N test (PSNT) times, respectively. Silking commonly occurred between Samplings #6 and #7 and the average of PAN measured at these two samplings is indicative of the pool of N that has been available through the rapid growth phase. The PAN at maturity (taken as the average of the last two measurements) includes the total N that has been utilized by the above ground part of the crop and the residual mineral N remaining in the soil.

The year effect in the VCEs would be expected to primarily reflect the effect of variation in weather between years. However, additional factors that may have contributed to the year effect include a progressive decline in the residual organic N associated with manure applied before the beginning of the experiment, and changes in the amount of residual N arising from the reduction in the amount of fertilizer-N applied to the barley in 2000 to 2003 from that applied in 1996 to 1999. Values of SMN at the time of preside-dress N soil testing and yield response to N would be expected to be most sensitive to variation in residual N related to previous manure or fertilizer treatments. Analyses of SMN at Sampling #3 and the yield response to N did not demonstrate any significant trend from 1997 to 2003, suggesting there were no long-term effects of the manure applications before initiating the study. These results were compatible with observations that the effects of manure on yield response of corn to a single application of manure were small by even the second year after application (Beauchamp, 1987). Values of SMN at Sampling #3 in the period 1997 to 1999 were not significantly greater than values in the period 2000 to 2003 suggesting that changes in the N fertilization of the barley had a negligible effect on the residual N. Consequently we assumed that the year effect was largely related to weather.

To provide an analytical framework for quantifying the effects on PAN of weather, soil and weather by soil interactions on PAN, the accumulation of PAN through the season was described by either a single first order kinetic function (Stanford and Smith, 1972), in which time was replaced by thermal time expressed as CDD, i.e.,

Formula 1[1]
or a linear function of CDD, i.e.,

Formula 2[2]
where a, b, c and d are constants. The choice of function and the use of CDD rather than day of year (DOY) were based on an evaluation (Dharmakeerthi et al., 2005) of four different functions (linear, single first order kinetic, Gompertz, and Logistic) to describe the temporal variation of PAN. Dharmakeerthi et al. (2005) found that the single exponential function generally accounted for more of the variability in PAN than the other functions when nonlinearity occurred and that a better fit was obtained when CDD was used rather than time (i.e., DOY).

Equations [1] and [2] embody several factors related to soil water content that need to be acknowledged when using the function for interpreting the effects of weather on PAN at different locations in the landscape. The first is a simple consequence of curve fitting; short-term variation in PAN due to factors influencing N mineralization, such as fluctuations in soil water content, are integrated into the shape of the function and therefore into interpolated values of PAN. The second relates to the use of CDD based on air temperature and its relation to soil temperature. The response of soil temperature to variation in air temperature is related to albedo and water content and therefore spatial variation in these characteristics would be expected to contribute to variation in soil temperature across a landscape. Although Dharmakeerthi et al. (2005) found that the spatial variation in soil temperature on this site beginning in early July was small (1–2°C), any effects of soil conditions on the relation between soil and air temperatures are incorporated into Eq. [1] and [2]. The third involves the relation between CDD, based on soil temperature, and rates of mineralization of organic forms of N. Honeycutt and Potaro (1990) concluded that soil thermal units expressed as cumulative degree days above 0°C provided a "mathematically simple, pragmatic approach" for predicting N mineralization. Although the net N mineralization was predicted adequately with soil thermal units for different levels of temperature x moisture combinations at soil water potentials within the –0.03 to –0.01 MPa range, Doel et al. (1990) speculated that this relationship may not be valid for prolonged drier conditions. These three factors, singly or in combination, would be expected to result in the shape of curves generated by fitting PAN data to Eq. [1] or [2] varying with soil and rainfall conditions.

The combined effects of soil and weather on PAN (Objective 2 and 3) were assessed by (a) identifying the stage during the growing season in which PAN was most strongly correlated to yield in the 0N treatment, (b) determining the soil properties having the strongest impact on PAN at that stage, (c) establishing the period during the growing season in which the CRF accounted for most of the variability in PAN and then (d) using multiple regression analyses to describe the dependence of PAN at a specific CDD on soil properties, rainfall and their interaction. Backward stepwise multiple regression analyses were used with OC, OC2, CRF, CRF2, and all their interaction terms in the initial model. During the analyses higher order interaction effects that were not significant at P < 0.05 were removed one at a time from the set of variables and the stepwise procedure rerun. This procedure was repeated until only significant terms were retained in the model. Stepwise multiple regression analyses were conducted using the STATISTICA (data analysis software system), version 6 (StatSoft, Inc., 2003). Goodness of fit was assessed using the coefficient of determination (R2), root mean square error (RMSE) and the normalized root mean square error (NRMSE), i.e.,

Formula 2
Analyses of variance were conducted using PROC GLM, and nonlinear regression analyses were conducted using PROC NLIN program of the SAS software package (SAS Institute Inc., 1996, Cary, NC).


   RESULTS AND DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
CONCLUSIONS
 REFERENCES
 
Weather
The temporal variation in CDD and CRF are given in Fig. 1 and 2, respectively. The weather exhibited considerable variability between 1997 and 2003. For instance, at DOY 200, the CDD varied from 1388 in 1997 to 1767 in 1998 (Fig. 1). The CRF exhibited even greater variability ranging from 192 mm in 1998 to 430 in 2000 on DOY 200 (Fig. 2). The distribution of CRF in the different CDD periods for the years 1997–2003 and the average for the 15-yr period 1986–2000 are given in Table 1. Rainfall in 1998 was below the 15-yr average in each of the four periods whereas it was well above average in 2000 in the first three periods.


Figure 1
View larger version (14K):
[in this window]
[in a new window]
  		Fig. 1. Seasonal variation in the accumulation of degree days.

 

Figure 2
View larger version (14K):
[in this window]
[in a new window]
  		Fig. 2. Seasonal variation in the accumulation of rainfall.

 

View this table:
[in this window]
[in a new window]
  		Table 1. Distribution of rainfall through different periods in the growing season, defined by cumulative degree days (CDD).

 
Plant Available Nitrogen
The variation in SMN, plant N (PN) and PAN with CDD in the 0N treatment is illustrated in Fig. 3 and the four critical sampling periods noted. Data were averaged across reps, management treatments, landscape positions, and all years except 2000. Data for 2000 were not included because excessive rainfall delayed seeding by about 4 wk and therefore the sampling times coincided with different dates in 2000 compared to other years. An interesting feature of this figure is the plateau in PAN that occurred about the time of silking. This feature was most pronounced in data from 1998 and 1999. Although Karlen et al. (1988) noted a decline in total N uptake around the time of silking, the total N in the plant in our study increased continuously through this period (Fig. 3). Therefore the plateau in PAN was due to the rapid decline in SMN rather than a decrease in N uptake. The soil water content declined from the fourth to the seventh sampling and the possible implications for the decline in SMN are considered in more detail later when data from different years are compared.


Figure 3
View larger version (17K):
[in this window]
[in a new window]
  		Fig. 3. Accumulation of soil mineral N (SMN), total shoot N (PN), and plant available N (PAN) in the 0N plots over the growing season (average across reps, management treatments, landscape positions and all years except 2000). Bars describe standard deviation among years of average across reps, management treatments and landscape positions.

 
Descriptive statistics for PAN measured at preplant, preside-dress, silking and maturity of maize as well as SMN at maturity are given in Table 2 for all years, excluding 2000. The large variability in PAN at the different sampling times illustrate the combined temporal (seasonal) and spatial (location) variability. The coefficients of variation (CV) in PAN (not presented) were largest at preplant and preside-dress and smaller at silking and maturity. Values of PAN at each sampling stage and SMN at maturity were always significantly larger following the barley + red clover-CT treatment than the other two treatments reflecting the contribution of N from red clover residues. Values of PAN were only greater in the barley-CT treatment than in the barley-NT treatment at preside-dress and silking times.


View this table:
[in this window]
[in a new window]
  		Table 2. Descriptive statistics for plant available N (PAN) and soil mineral N (SMN) at 0–30 cm assuming bulk density of 1.3 Mg m–3 for all years (excluding 2000) and landscape positions and each management treatment in the 0N treatment.

 
Factors Contributing to the Variance in Plant Available Nitrogen
Variance component estimates (VCEs) of the main effects and their interactions along with the relevant CVs are given in Table 3 with the order of the terms coinciding with the order appearing in the analyses. Year, management treatment and landscape position accounted for most of the normalized variance and they were larger than the interaction terms (Table 3). Year, management treatment, and landscape position were significant at all sampling times except at preplant where the treatment effect would not have been manifested and therefore was not significant (Table 3). The CVs for the year and treatment effects decreased from preside-dress sampling to maturity but increased for the position in the landscape effect. The data are compatible with the data obtained by Sogbedji et al. (2001) who found that year and the interaction effect between year and spatial variability (drainage class) contributed significantly to economically optimum N rates.


View this table:
[in this window]
[in a new window]
  		Table 3. Variance component analysis of plant available N at different sampling stages in the 0N treatment for 1997, 1999, 2001, and 2003.

 
A comparison of VCEs (Table 3) indicates that position in the landscape is an important component of the variation in PAN suggesting a potential opportunity for using site-specific N management on this landscape. However, the combination of year and year by position by treatment has an effect that is similar in magnitude to position alone. This would imply that the potential benefits of site-specific management would only be fully captured if N management could be adjusted annually in response to changing weather conditions, especially those early in the growing season. Such adjustments would be facilitated by quantifying the effects of weather using readily measurable weather variables.

Impact of Air Temperature and Rainfall on Plant Available Nitrogen
To assess the impact of air temperature on PAN, PAN was first related to CDD using Eq. [1] or [2]. Of the 105 data sets (7 yr, 5 landscape positions, 3 management treatments), linear behavior was exhibited by 40 (38%) of the data sets (as indicated by higher R2 and coefficients with a higher P value than for the exponential function). Of these, it was most commonly found through the period 2000 to 2003 (38 data sets), in the summit, footslope, and toeslope positions (28 data sets), and in the barley NT and barley CT treatments (38 data sets). The variation in the shape of the curves is illustrated in Fig. 4 for the shoulder and footslope slope positions in 1999, and 2001 for the barley-NT treatment. The OC contents were 13.7 and 11.1 g kg–1 in the plots at the shoulder slope position, and 22.4 and 21.9 at the footslope position in 1999 and 2001, respectively. Although the PAN that had accumulated by the end of the season was similar in both years at a given position, differences were obvious early in the season and these differences were reflected in the shapes of the curves. The exponential-shaped curves for PAN in 1999 (Fig. 4a) contrasted with the more linear behavior exhibited in 2001 (Fig. 4b). The difference in shape of the curves between years suggests that the accumulation of PAN was influenced by other factors in addition to air temperature.


Figure 4
View larger version (24K):
[in this window]
[in a new window]
  		Fig. 4. Variation between years in the relation between Plant Available N (kg ha–1) and Cumulative Degree Days (CDD) illustrated for the backslope and footslope positions in 1999 (Fig. 4a) and 2001 (Fig. 4b) for the barley-NT treatment.

 
Differences in the accumulation of PAN early in the season were most obvious between the first and third sampling. For instance, PAN increased by 46 kg ha–1 between these two samplings at the footslope position in 1999 but decreased by 7 kg ha–1 in 2001. The loss in PAN in 2001 coincided with larger rainfall from 200 to 700 CDD in 2001 compared to 1999 (Table 1). The water contents of soil are normally higher at this time of the growing season than later in the season, and larger rainfall in 2001 may have contributed to increased denitrification or leaching or to slower increase in soil temperature. These effects of larger early season rainfall would be expected to be accentuated in the positions in the landscape that are the wettest. The gravimetric water contents, averaged over the first three samplings in 2001, were 0.19 and 0.24 g g–1 at the 0–0.30 m depth for the backslope and footslope slope positions, respectively. On the wetter footslope position, PAN decreased over the first three samplings by 7 kg ha–1 compared to a small increase of 6 kg ha–1 on the dryer backslope position (Fig. 4b). A linear variation in PAN with time was most common on the wettest positions in the landscape. A linear relation occurred more frequently in the barley-CT and-NT treatments than in the treatment that included red clover. The exponential relation that dominated the relation between PAN and CDD in the barley + red clover-CT treatment may have reflected an overriding effect of the labile source of N in the legume.

Differences in the accumulation of PAN mid-season were most obvious between the fifth and seventh sampling. The nature of the plateau in PAN noted with respect to Fig. 3 varied among years (Fig. 4) with the effect being more pronounced in 1999 than in 2001. As noted with respect to Fig. 3, the PN increased through this period and the decrease in the rate of accumulation of PAN in Fig. 4a was due to an apparent disproportionately large decrease in SMN. The data from the backslope and footslope positions (Fig. 4a) were representative of the other positions in the landscape in 1999. Among all years, data from 1998 were most similar to those in 1999. Soil water content declined from the fourth to the seventh sampling (see Fig. 1, Dharmakeerthi et al., 2004) thereby introducing the possibility of a drought-driven decrease in the rate of N mineralization. However, the water content at the seventh sampling fell to a slightly smaller value in 2001 than in 1999 suggesting that the difference in the behavior of PAN in 1999 and 2001 might not be attributed to drought-driven differences in the rate of N mineralization. An apparent large decrease in SMN would also arise if there was a disproportionately large increase in N immobilization or in the amount of N retained in root tissue. An increase in N immobilization would be dependent on a rapid increase in the amount of labile C (arising perhaps from an increase in root mortality or the production of root exudates). An increase in root N would be dependent on changes in the proportion of N in the roots versus the shoots. Although neither of these possibilities can be assessed, the SMN at the fifth sampling was much larger in 1999 than in 2001 (101 and 47 kg N ha–1, respectively when SMN is averaged across the two landscape positions given in Fig. 4), and the SMN represented a larger proportion of PAN in 1999. This may have contributed to the apparent decrease in SMN being much more evident in 1999. Further research is necessary to elucidate this phenomenon.

Integrating Effects of Soil and Weather Factors to Describe Variation in Plant Available Nitrogen
Since the effects of weather on PAN varied through the season, efforts to quantify the effects of weather on PAN were focused on values of PAN at a time in the growing season that appeared to have the largest effect on grain yield in the 0N treatment. The best fit relations between PAN and CDD (Eq. [1] or [2]) were used to calculate values of PAN at 350, 700, 1350, 2000, and 2600 CDD and these values were then correlated with yield. The correlations (not given) improved as the growing season progressed, with the strongest correlations found between yield at 0N and PAN at 2000 and 2600 CDD. Values of PAN at 2000 CDD were then used as the dependent variable in subsequent regression analyses with soil characteristics and CRF in different periods used as the independent variables.

The soil properties exhibiting the greatest spatial variation on this site were OC content and soil water content (Dharmakeerthi et al., 2005) and these variables were strongly correlated. Dharmakeerthi et al. (2005) found that OC content was positively correlated to PAN at different stages in the growing season in 1999–2001 and that the relation between PAN and OC changed from a linear relation in the early part of the growing season to a quadratic relation later in the season. The positive correlation of organic matter with N mineralized has been noted in other studies (Qian and Schoenau, 1995; Barrett and Burke, 2000; Deng and Tabatabai, 2000), although Schmidt et al. (2002) did not find that locations representing large OC contents coincided with areas that required the least amount of N to achieve maximum yield. Values of OC were selected for use in subsequent analyses to account for the influence of spatial variation in soil properties on PAN at 2000 CDD.

Multiple regression analyses were employed to relate PAN at 2000CDD in a management treatment to OC and CRF separately for the periods 200 to 700, 700 to 1350, and 1350 to 2000 CDD using the data from all years. Backward stepwise multiple regression analyses were used with OC, OC2, CRF, CRF2, and all their interaction terms in the initial model. The CRF for the period 200 to 700 CDD was the only CRF term that was significant and appeared in the regression equations for all management treatments (Table 4).


View this table:
[in this window]
[in a new window]
  		Table 4. Stepwise regression analyses relating plant available N (kg ha–1) at 2000 cumulative degree days (CDD) to soil organic C (OC) (g kg–1) and cumulative rainfall (CRF) (mm) over 200 to 700 CDD.

 
The greater importance of the CRF from 200 to 700 CDD may reflect the effect of rainfall in this period on the initial rate of accumulation of PAN, as illustrated in Fig. 4. The relatively smaller importance of CRF in later periods is attributed to the range in soil water contents through these periods in the context of the relation between net N mineralization and water content in these soils. We have observed (Drury et al., 2003) that there is a range in water content within which the net N mineralization is maximum and relatively independent of water content, and have defined this as the nonlimiting water range (NLWR). Dharmakeerthi et al. (2005) noted that the soil water content, averaged over all sampling times in the season, fell within the NLWR at all positions in the landscape in 1999 and 2000; although they noted that the water contents may fall outside of the range for some periods in the season. The relation between net mineralization and soil water content (expressed as percent water filled pore space, WFPS) is illustrated in Fig. 5 for the soil from 0 to 0.20 m taken from the toeslope position on this site (adapted from Drury et al., 2003). The WFPS at –0.01 and –1.5 MPa were 82 and 39%, respectively. The NLWR for this soil extended from 53 to 78% WFPS. The contrasting shapes of the curve at the dry and wet end of the curve are particularly relevant to the impact on PAN of the CRF in the different periods. At the dry end, net mineralization declined slowly as the water content fell from the lower limit of the NLWR to that at the permanent wilting point. However, at the wet end of the curve, the net mineralization declined precipitously when the water content exceeded the upper limit of the NLWR. The water content at field capacity was greater than that at the upper end of the NLWR. The soil water content, through the periods of greatest plant growth and transpiration (700 to 1350 and 1350 to 2000 CDD), would be expected to range from field capacity to permanent wilting point. This expectation was confirmed in an examination of actual water contents through this period. The mean and standard deviation of the WFPS in the period 700 to 1350 CDD (fourth and fifth sampling; Fig. 3), considering data for all landscape positions, management treatments, and years (except 2000), were 54 and 10%, respectively. Similarly the corresponding mean and standard deviation of the WFPS for 1350 to 2000 CDD (sixth and seventh sampling) were 43 and 8%, respectively. On the basis of Fig. 5, minimal variation in net mineralization would be expected, with the range in water contents encountered in the 700 to 1350 and 1350 to 2000 CDD period, and therefore variation in rainfall would not be expected to have a large impact on PAN. Therefore, it is not surprising that among the three periods the CRF in the 200 to 700 CDD period was most strongly correlated with PAN at 2000 CDD. The strong correlation between PAN at 2000 CDD and yield in the 0N treatment might be expected to result in early season rainfall influencing yield response to N under similar soil and climatic conditions. This expectation is confirmed in studies in New York State (Sogbedji et al., 2001).


Figure 5
View larger version (12K):
[in this window]
[in a new window]
  		Fig. 5. Variation in net mineralization of N with soil water content (expressed as water-filled pore space) of soil from the toeslope position (adapted from Drury et al., 2003).

 
The models in Table 4 describe strongly nonlinear behavior in PAN with CRF in the period 200 to 700 CDD. The model for the barley-NT treatment is illustrated in Fig. 6 . The nonlinear behavior was greatest on sites with sizable OC content and large decreases in PAN coincided with high rainfall. The sites with large OC content coincided with those positions in the landscape with highest soil water content and were generally the foot- and toeslope positions. Loss of N through denitrification and leaching would be expected to be most common under these conditions.


Figure 6
View larger version (65K):
[in this window]
[in a new window]
  		Fig. 6. Illustration of the variation in Plant Available N at 2000 CDD with organic carbon content and cumulative rainfall in the period 200 to 700 CDD described by the regression equation in Table 4 for the barley-NT treatment.

 
Assessment of Models Using Independent Data Sets
The models in Table 4 were assessed using data obtained from two other experiments on similar textured soils at the Elora Research Station and in which the crop rotations were similar to those in this study. The first was a long-term rotation study established in 1980 that included a barley/barley/maize/maize and a barley + red clover/barley + red clover/maize/maize rotation under conventional tillage. Additional details on this site were provided by Yang and Kay (2001). Data were obtained from the first year of maize in both rotations. The second study had been established in 1996 on a simple slope similar to that in the current study and included a barley/maize rotation under conventional tillage. Data were obtained from the maize phase of the rotation. Data collected in 1998 were obtained from both studies. Values of PAN predicted from models in Table 4 were similar to those calculated at 2000 CDD from measured values (Fig. 7 ) although the model slightly overpredicted PAN on plots on the simple slope. The reasons for this overprediction were not identified. The assessment has the obvious shortcoming that it was based on data from a year in which the model was developed (similar data were not available for years outside of the range in which the models in Table 4 were based). However, setting aside that limitation, the correspondence between the observed and predicted values suggests that the model should apply to soils in which (a) the shape of the net mineralization versus water content curve is similar to that in Fig. 5, that is, net mineralization exhibits greater change due to variation in water contents near field capacity than between field capacity and permanent wilting point, and (b) drainage conditions result in the soil water content through the growing season remaining below the upper limit of the NLWR and therefore the loss of N by denitrification is small.


Figure 7
View larger version (13K):
[in this window]
[in a new window]
  		Fig. 7. Comparison of Plant Available N at 2000 cumulative degree days observed on two research sites in 1998 at the Elora Research Station with that predicted from the regression equations in Table 4.

 
Use of Models with Historic Weather Data to Assess Temporal Stability in Spatial Patterns of Plant Available Nitrogen
The nonlinear dependence of PAN on OC and CRF described by the models in Table 4 introduces the possibility that patterns in PAN vary among years, that is, the spatial patterns fail to exhibit stability. The frequency distribution of PAN at 2000 CDD over a 15-yr period was assessed by employing the models in Table 4 with weather data from the site from 1986 to 2000 inclusive). The CDDs were calculated for each day through the growing season in each year from 1986 to 2000 inclusive and then the CRF from 200 to 700 CDD calculated for each year. The resulting values of CRF were normally distributed with a mean, standard deviation, minimum and maximum of 90.3, 34.7, 38.8, and 154 mm, respectively (the CRF in 1 yr fell just outside of the range in CRF on which the models in Table 4 were based). Values of CRF for each year were substituted in the models to predict the PAN at 2000 CDD in soils with three different OC contents: 10, 20, and 30 g kg–1 for each year. The distribution in the predicted values of PAN throughout the 15-yr period at the three OC contents is illustrated in Fig. 8 for the barley-NT treatment. The range in the distribution in PAN was smallest for the soil with an OC content of 20 g kg–1 and greatest for the soil with 30 g kg–1. In 12 of the 15 yr, PAN increased with increasing OC content. However, in 1 of the 15 yr, PAN in soil with 30 g OC kg–1 was about the same as in soil with 20 g kg–1, and in 2 of the 15 yr, PAN in soil with 30 g OC kg–1 was smaller than that in soil with 20 g OC kg–1. These years had the largest CRF between 200 and 700 CDD and corresponded to points falling in the upper right hand corner of the surface described in Fig. 6. Thus, the spatial pattern in PAN might be considered to be generally stable in that PAN increased with OC in 12 of the 15 yr.


Figure 8
View larger version (25K):
[in this window]
[in a new window]
  		Fig. 8. Frequency distribution of Plant Available N at 2000 cumulative degree days predicted for each year through the period 1985–2000 for organic C contents of 10, 20, and 30 g kg–1 using the model for the barley-NT treatment (Table 4) and cumulative precipitation from 200–700 CDD.

 

   CONCLUSIONS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
REFERENCES
 
Year, management treatment, and landscape position accounted for most of the normalized variance in PAN at all sampling times, except at preplant where the treatment effect would not have been manifested and therefore was not significant. The main effects were larger than the interaction terms. Although position in the landscape was an important component of the variation in PAN, large effects arose from the combination of year, year by treatment, and year by position by treatment.

PAN increased with increasing CDD and showed the strongest correlation with yield in the 0N plots at 2000 and 2600 CDD. The variation in PAN due to spatial variation in soil characteristics and temporal variation in weather was described by the spatial variation in OC, the temporal variation in CDD and CRF, and the interaction between OC and CRF. Early season rainfall (200–700 CDD) had a more significant effect on PAN at 2000 CDD than rainfall later in the season. This may have been due to the temporal variation in soil water content and the shape of the net mineralization/water content curve for this soil. Net mineralization showed much greater sensitivity to water contents above field capacity (which would be encountered more often early in the growing season in a cool humid environment) than between field capacity and the permanent wilting point (which would be encountered later in the growing season).

At 2000 CDD, PAN was strongly nonlinearly related to OC and CRF in the 200 to 700 CDD period; PAN increased with OC at low CRF but exhibited a decline at large OC contents at high CRF. The decline was attributed to losses of N by leaching or denitrification under high rainfall in those parts of the landscape which had the largest OC and which were also the wettest. Changes in the relation between PAN and OC with early season rainfall introduce the possibility that the spatial patterns in PAN are not temporally stable. Application of the regression models to weather records (1986–2000 inclusive) indicated that the spatial patterns in PAN at 2000 CDD were generally stable and increased with OC in 12 out of 15 yr.

The analyses suggest that although there is a potential opportunity for using site-specific N management on this landscape, its potential benefits would only be fully captured if N management could be adjusted for different weather conditions early in the growing season.


   ACKNOWLEDGMENTS
 
The Ontario Corn Producers' Association, Agriculture and Agri-Food Canada, the Natural Sciences and Engineering Research Council of Canada, and the Ontario Ministry of Agriculture and Food provided financial support for this research. The Canadian Commonwealth Scholarship and Fellowship Program provided financial support for R.S. Dharmakeerthi to pursue studies leading to the Ph.D. degree at the University of Guelph. The assistance of Jim Ferguson with fieldwork and Ranee Pararajasingham with technical and data analyses is also gratefully acknowledged.

Received for publication February 1, 2005.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 





This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via ISI Web of Science (3)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Kay, B. D.
Right arrow Articles by Dharmakeerthi, R. S.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Kay, B. D.
Right arrow Articles by Dharmakeerthi, R. S.
Agricola
Right arrow Articles by Kay, B. D.
Right arrow Articles by Dharmakeerthi, R. S.
Related Collections
Right arrow Nutrient Management
Right arrow Plant Analysis

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Agronomy Journal Crop Science
Journal of Natural Resources
and Life Sciences Education