| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a Département des Sciences de la Terre et de l'Atmosphère, Université du Québec à Montréal, C.P. 8888, succ. Centre-Ville, Montréal, QC, H3C 3P8 Canada
b Laboratoire d'Hydrogéologie, Faculté des Sciences, Université d'Avignon, 33, rue Louis Pasteur, 84000 Avignon, France
c Direction de la recherche forestière, Forêt Québec, Ministère des Ressources naturelles (MRN), 2700, Einstein, Sainte-Foy, QC, G1P 3W8 Canada
d Centre de Recherche en Biologie Forestière (CRBF), Faculté de foresterie et de géomatique, Université Laval, Sainte-Foy, QC, G1K 7P4 Canada
* Corresponding author (larocque.marie{at}uqam.ca)
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONREFERENCES |
|---|
Abbreviations: AS, ammonium sulfate treatment • CV, coefficient of variation • RMSE, root mean squared error • SCU, S-coated urea
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONREFERENCES |
|---|
In many aspects, the cultivation practices used to produce bareroot seedlings in forest tree nurseries are similar to those employed in agricultural fields (use of fertilizers and pesticides, irrigation). Two important differences, however, are the very shallow rooting depths in forest tree nurseries (20 cm) and the recurrent fertilizer applications carried out during the growing season of forest tree seedlings. Moreover, most forest tree nurseries produce bareroot seedlings in well-drained soils which are highly susceptible to NO3 leaching. Many mechanistic models have been developed in recent years to simulate agricultural nutrient losses to the environment: SOIL-SOILN (Johnsson et al., 1987), DAISY (Hansen et al., 1991), LEACHM (Husston and Wagenet, 1992), RZWQM (USDA-ARS, 1992), and AGRIFLUX (Banton and Larocque, 1997) are just a few examples. To our knowledge, there are currently no models that specifically simulate the fate of N fertilizers under tree seedling crops. The one that comes the closest is the FORESTSR, a submodel of the SOIL-SOILN model (Eckersten, 1994), which simulates NO3 losses under 0- to 4-yr tree stands. Because of the context similarity, models developed for agricultural soils could be used to simulate the N dynamics in forest tree nursery soils, making them potential management tools for nurserymen.
The objective of this study was to determine if a model simulating the N dynamics in agricultural soils could be used to estimate soil inorganic N levels in forest tree nursery soils. The AGRIFLUX model (Banton and Larocque, 1997) was selected because it has been tested and has proven to be reliable in different agricultural field conditions (Larocque and Banton, 1995; Dupuy et al., 1997; Larocque et al., 1998; Lasserre et al., 1999). In the current study, this model was used to simulate the inorganic soil N contents under four N fertilizer treatments. The model was applied to a forest tree nursery (Saint-Modeste nursery, Forêt Québec, MRN) located in Quebec, Canada.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONREFERENCES |
|---|
|
|
|
The processes represented in the N cycle include fertilizer inputs (source, timing, and amount of inorganic and organic N), mineralization, nitrification, denitrification, and N uptake by seedlings. Nitrate in the soil solution can be mobilized with runoff, drainage, and leaching water. Biochemical processes are linked to the soil temperature and water content through the reaction rates. Both soil temperature and water content influence rates of mineralization, nitrification and denitrification. Most mathematical representations of the different soil N cycle processes are based on those used by Johnsson et al. (1987).
In AGRIFLUX, the seedling N uptake curve is represented with a normal distribution curve truncated at plus and minus three standard deviations. The start and end of the curve correspond to the beginning of the growing period and to the end of seedling N uptake. The area under the curve is the total seedling N requirement. The beginning and the end of the growing period and total N requirements are easily determined from a seasonal growth curve of the total dry mass of seedlings (Langlois and Gagnon, 1993). This nutritional approach has been tested with field measurements of seedling N uptake (Banton and Marquis, 1996). Although it has not been tested for water uptake, the approach is assumed to be applicable to this process as well. Evapotranspiration is mostly related to seedling water uptake and occurs during a limited time each year. The model user provides monthly potential evaporation data and the model distributes this value over each month. The amount of water corresponding to the combination of potential evaporation and seedling water uptake is extracted from the soil (as a function of evaporation and rooting depth) if enough water is available.
The Experimental Site
The field experiment was designed to measure the impact of different N fertilization treatments on the availability of inorganic N (NH4 and NO3) in a forest tree nursery soil. Two sources of N were compared: NH4 as AS (21-0-0) and urea as SCU (38-0-0), both applied at different frequencies. The selected experimental site (Saint-Modeste nursery) is a provincial forest tree nursery (Forêt Québec, MRN) located in the municipality of Saint-Modeste ( 47°50' N. lat., 69°30' W. long.), Quebec, Canada. The soil texture of Seedling Beds 4 and 9 in the selected nursery Block 130 was a sandy loam. This soil is classified under the orthic humo-ferric podzol and orthic regosol groups. It corresponds to a loamy, mixed frigid Typic Haplorthod. The free watertable was located at an average depth of 12 m. In the spring of 1993, Seedling Bed 4 received the equivalent of 170 kg N ha-1 from a compost made of horse (Equus caballus) manure and carrot [Daucus carota L. ssp. (Hoffm.) Arcang.] (100 m3 ha-1), whereas no compost was used in Seedling Bed 9. This amendment was intended to increase the soil organic matter content in Seedling Bed 4 (see Table 2). From 1993 to 1995, bareroot black spruce seedlings [Picea mariana (Mill.) B.S.P.] were grown in both seedling beds. At the end of the 1993 growing season, the black spruce seedlings were 3 yr old and were 2 + 1 stock (the first number represents the number of years in the seedbed, and the second number corresponds to the number of years after transplantation in another cultural block). Consequently, 4-yr-old seedlings (2 + 2 stock) were produced in 1994 and 5-yr-old seedlings (2 + 3 stock) were produced in 1995. Fifteen irrigation applications were performed at a rate of 5 mm per 30-min period from the end of May to the middle of August 1993. Irrigation was applied using sprinklers whenever necessary to satisfy the water requirements of the 3-yr-old seedlings (2 + 1 stock) and to maintain a soil water content close to field capacity. No irrigation was applied in 1994 and 1995.
In May, 1994, a completely randomized design with four N fertilizer treatments (T1, T2, T3, and T4) and three replicates (plots) was set up in each seedling bed. Plots measured 2 by 1.5 m2 (the normal width of cultivation plots in forest tree nurseries is 2 m). For all N fertilizer treatments, a total of 186 kg N ha-1 was applied during the 1994 growing season to meet the growth requirements of the 4-yr-old seedlings (Table 3). The AS treatment (T1) was applied as a fertilizer solution in each plot: nine applications of 300 mL plot-1 (297 g L-1). The SCU treatments consisted of broadcast granules over the plots applied in one application per plot (T2), two applications (T3), or three applications (T4) (Table 3). In 1995, the 5-yr-old seedlings (2 + 3 stock) received no fertilization.
Soil NO3 and NH4 concentrations were determined four times in 1994 (June 9, July 14, August 31, and October 18) and twice in 1995 (June 8, October 4). At each sampling date, one bulk soil sample was collected from two depths (0–20 cm for the root zone and 20–40 cm immediately below the root zone) in each of the 3-m2 plots. A total of 24 soil samples were therefore collected from each seedling bed at each sampling date. The soil samples were sieved at 2 mm and analyzed for inorganic N concentration. Inorganic N was extracted with 2 M KCl and NH4 and NO3 concentrations (mg N kg-1) were determined by flow injection analysis (Tecator FIAstar 5020 Analyzer, Tecator Inc., Herndon, VA).
To estimate the seasonal seedling N uptake in each seedling bed, 18 seedlings were harvested from each N fertilizer treatment (six seedlings per plot x 3 plots) at the end of the 1994 and 1995 growing seasons. This sampling frequency was designed to limit the impact of harvests on the seedling density (
60 seedlings m-2). At each harvest, the growth parameters of the seedlings were measured (shoot height, root-collar diameter, and shoot [stem, leaves] and root dry weights) and their N contents were analyzed. Total N was determined separately in shoot (stem and leaves) and root tissues following digestion in concentrated H2SO4, H2O2, and Se in digestion tubes maintained at 370°C. The N contents in the digest solutions were measured by colorimetry using a continuous flow spectrophotometer (model QuickChem 8000, Lachat Instruments, Milwaukee, WI).
Undisturbed soil samples were collected once in 1999 to determine hydrodynamic properties and bulk density. The samples were collected at 16 random locations from Block 130 and from adjacent blocks with sandy loam soils. At each location, undisturbed samples were taken at depths 5 to 15, 25 to 35, and 55 to 65 cm. The samples were analyzed in the laboratory to determine saturated hydraulic conductivity using a constant head method (Klute and Dirksen, 1986). The saturated samples were weighed and put to drain on a tension table (Topp et al., 1993) for 3 d and weighed again to calculate drainage porosity. The drained samples were oven-dried at 105°C for 24 h and weighed again to determine dry weight. Bulk density was obtained by dividing sample volume by dry weight. Soil sampling for organic matter content was performed in May 1991. A composite sample was collected from random locations in the 0- to 20-cm layer of both seedling beds. The organic C content was determined using the Walkley-Black method (Nelson and Sommers, 1982). Table 2 presents the soil characteristics.
Statistical Analysis
A statistical analysis was conducted to determine if all of the experimental conditions should be simulated separately or if some scenarios could be combined. An analysis of variance was used to determine if the effect of N fertilizer treatment and seedling bed on soil inorganic N concentrations was significant. For each soil sampling date, the GLM procedure of the statistical package SAS (SAS Institute Inc., 1990) was used to perform an analysis of variance (
= 0.1) of the soil inorganic N concentrations in response to N fertilizer treatment.
Among the different methods that can be used to establish the accuracy of a model for the prediction of measured values, the Root Mean Square Error (RMSE) was chosen as a deviance measurement (Willmott et al., 1995). The hypothesis that differences between measured and simulated soil N are independent and normally distributed with a mean of zero was tested by performing a t–test on the average difference between measured and simulated results (average residual). The average of absolute residuals is used to quantify the average simulation error.
Model Parameters
The soil profile was divided into four layers. The first layer (0–20 cm) corresponds to the rooting depth and to the first sampling depth. The second layer (20–40 cm) lies immediately under the rooting depth and corresponds to the second sampling depth. Two other layers (depth of 30 cm each) were added to simulate NO3 migration under the seedlings. The simulations began in November 1991 and ended in November 1995. This extended period was used to minimize the impact of initial conditions. The parameters required in AGRIFLUX are summarized in Table 1.
Because AGRIFLUX is a stochastic model, a distribution can be used for all its input parameters. Normal distributions with a coefficient of variation (CV) of 20% were used for all the parameters except hydraulic conductivity, for which a lognormal distribution with a CV of 100% was used. Use of the normal distribution was based on field measurements of some parameters and extended to all parameters as a default value. In previous work (Larocque and Banton, 1995), a CV of 10% was observed to be too small to represent the variability of the phenomenon, and a larger value was therefore used in the current study. The parameters were used as measured or found in literature and were not calibrated. One hundred Monte-Carlo simulations were performed for each N fertilizer treatment.
Sensitivity Analysis
A sensitivity analysis (McCuen, 1973) was performed to identify the most influential parameters, which need to be determined with the greatest accuracy. The relative sensitivity coefficient, Sr (ratio between the relative variation of the studied result, 
F/F, and the relative parameter variation, X/X), provides a direct comparison of the influence of the different parameters. The sensitivity analysis was performed using all the parameter values required to describe one of the plots receiving AS (T1) as a reference scenario. The parameters were modified within a plausible variation interval (–50 to +50%) and were analyzed one by one. No cross-correlations were investigated. Three model results were used in the analysis: the simulated soil NO3 and NH4 contents averaged over the simulation period (measured model output), and the potential NO3 leaching out of the root zone (an unmeasured result).
|
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONREFERENCES |
|---|
|
|
|
Model accuracy was measured in different ways (Table 6). The RMSE for soil inorganic N content is smallest for T1 and greatest for T3. Therefore, the model simulates the soil N response to fertilizer treatment better with split inputs of AS over the growing season than the N fertilizer treatment with one to three inputs of SCU. The average residuals calculated for the four fertilizer treatments are small and not significantly different from zero (t-test, p < 0.05). This suggests that the model is not biased when data are pooled by fertilizer treatment. The RMSE is greater in the 0- to 20-cm layer than in the layer below the root zone and is similar for both seedling beds. The average residuals calculated for the first soil layer are significantly different from zero at p < 0.05. This suggests that the model may overestimate the soil inorganic N contents in the first soil layer. This could be because of underestimates of simulated N uptake by seedlings or simulated N losses through runoff. Losses by volatilization are not likely, at least for the SCU treatment. The model accurately simulates the inorganic N in the layer below the roots. This result is interesting because the inorganic N found in this layer is likely to find its way (at least partially) to the watertable. When pooling all the data, the absolute residual indicates that the average simulation error for the entire experiment is 25.2 kg N ha-1. The t-test performed on these residuals shows a slight global overestimation of soil inorganic N levels at 
= 0.05. Field measurements show that inorganic soil N contents can vary significantly. For a given sampling date, measured soil inorganic N varies from 0.5 to 441 kg N ha-1 in the 0- to 20-cm layer, and from 0.13 to 245 kg N ha-1 in the 20- to 40-cm layer. The greatest difference occurred in June 1994 and the smallest in October 1995. Considering this variability, the absolute average simulation error appears to be reasonable.
|
|
Sensitivity Analysis
The results of the sensitivity analysis (Table 7) show that most of the parameters tested have a |Sr| of <0.5. This means that variation in the model results is <50% of the variation of the input parameter.
|
The N content of the inorganic fertilizer (FER_N) has the most influence on the potential NO3 leaching out of the root zone, but has much less impact on the soil NO3 and NH4 content. This is explained by the NO3 increase following a large application of inorganic fertilizer. Since the amount of N required by the seedlings remains constant, the excess NO3 leaches rapidly downward because of high soil permeability and has little effect on soil NO3 and NH4 contents. In light of this sensitivity analysis, it can be expected (at least in the model) that the more frequent, smaller applications (T1) are less conducive to NO3 leaching to the groundwater. The high Sr value for this parameter does not preclude the use of the model because the N input from fertilizers is generally known.
Parameters related to the transformation of the organic matter (LITEFF, LITHF, HUMK, ORG_MAT, and CNORG) have a relatively large influence on inorganic N contents and NO3 leaching. This finding is important since these parameters are usually unavailable and must be obtained from literature. Moreover, it is known that the parameters responsible for organic matter transformation are interrelated; an increase in one parameter can decrease the influence of another parameter. These interactions were not tested in the current study, but have been reported by Larocque and Banton (1996). The seedling N uptake (PL_N) also has a significant impact on all three model output variables. This parameter is relatively well known and can be measured in the field.
The results of the sensitivity analysis also show that drainage porosity (DR_POR) has a significant influence on soil NO3 and NH4 contents. On the other hand, the saturated hydraulic conductivity (SATK) has little impact. These findings can be explained by the Darcy-Richards approach used in the model. This robust representation puts less emphasis on hydraulic conductivity (which is highly variable in the field and difficult to determine precisely) and more weight on the drainage porosity, a parameter that is easier to measure. These results are indications of the robustness of the AGRIFLUX model.
All of the other parameters have less influence on the results. For example, seedling water uptake has a Sr < 10% for all three response variables. This information is particularly interesting for nurserymen since seedling water uptake is very difficult to measure and those data are usually unavailable.
|
CONCLUSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONREFERENCES |
|---|
Results show that the fate of soil inorganic N was relatively well simulated for all N fertilizer treatments; the temporal changes in inorganic N content was well simulated overall; the decrease in N with soil depth was accurately represented; the low inorganic N contents in 1995 were well reproduced. The N fertilizer treatment with split applications (AS) was better simulated overall than the three other treatments (SCU) and also shows the least variability in soil inorganic N contents. This result is interesting since split N applications are among the current fertilization trends in forest tree nurseries. The inorganic N contents were slightly overestimated globally and in the root zone (0–20 cm). They were more accurately simulated in the layer below the roots (20–40 cm). Considering the variability of measured soil N levels, the overestimation appears to be relatively small and the results still provide a usable estimate of N availability to seedlings. The accurate simulation of inorganic N below the root zone is an indicator of a good simulation of NO3 leaching to the groundwater. The average absolute error of the model output is small compared with the measured variations of soil inorganic N contents.
Sensitivity analysis showed that most parameters used in the model had relatively little influence on the results. The most influential parameters were either available or can be measured easily. This is a sign of the stability of the model and indicates that the process representations are well-balanced in their complexity and reliability. With AGRIFLUX, processes as well as parameters, are easily adapted to forest tree nurseries because of the similarity between the two simulation contexts, and also because of the mechanistic representation of processes in the model. A model using empirical process representations might have proven less adaptable and less useful in a forest tree nursery environment. The high variability measured confirms the importance of using a stochastic model that takes parameter variability into account. It also emphasizes the need for field measurements of parameter variability in different soils and under different cropping conditions.
This study shows that a model like AGRIFLUX can be used satisfactorily in forest tree nurseries to simulate the fate of N fertilizers used in bareroot seedling production. These results are applicable in a Nordic climate with black spruce seedlings and similar coniferous seedlings, in a naturally well-drained soil and under limited irrigation (typical conditions in the forest tree nurseries of Eastern Canada). Different conditions could modify the soil N dynamics, but the use of a mechanistic model reduces the risk of an unreliable simulation.
This study leads the way to the use of NO3 leaching models, like AGRIFLUX, in the management of soil inorganic N in forest tree nurseries. Such a model could easily be adapted for management purposes by using default values for a number of parameters. To ensure reliability, the most influential parameters, which are easily obtained, could remain user-defined. The parameters that are not available or have little influence on the results could be set at standard values. This simplified simulation tool could be used by nurserymen to evaluate the inorganic N content of the soil, allowing them to adjust the amount of inorganic N fertilizers to minimize NO3 leaching while providing the necessary nutrients to the seedlings.
CNORG C/N ratio of microorganisms and humified productsDEN_MAXMaximum denitrification depth (m)DENHSHalf-saturation constant in the function for the NO3 concentration effect on denitrification (mg N L-1)DENPOTPotential rate of denitrification (g N m-2 d-1)DR_PORDrainage porosity (m3 m-3)EV_DEPMaximum evaporation depth (m)EVAPPotential evaporation (mm)FER_NQuantity of N added as inorganic fertilizer (kg N ha-1)FER_RELDays required for granule fertilizer to be released in the soil (d)GROWTHLength of the growing season delimited by the start and end of plant uptakeHORIZNumber of soil layersHUMKHumus mineralization rate (d-1)IRRIG Amount of water added through irrigation (mm)LITEFFEfficiency of the internal synthesis of the microbial biomassLITHFLitter C humification fractionLITKLitter decomposition rate (d-1)MELTRate of snowmelt (m °C-1 d-1)NITKNitrification rate (d-1)NITRNO3/NH4 ratio used in the nitrification functionORG_MATOrganic matter content of the soil (%)PL_NSeedling N uptake (mg m-2)PL_WATSeedling water uptake (mm)PRECMonthly precipitation (mm)ROOTSMaximum rooting depth (m)SATKSaturated hydraulic conductivity (cm s-1)SNOWFraction of snow left on the soil at end of winterTEMPMonthly air temperature (°C)TEMP_MINDay when the air temperature is minimumUPMAFraction of available mineral N for immobilization and plant uptakeWETNumber of wet days in one yearWILTSoil wilting point (m3 m-3)
|
ACKNOWLEDGMENTS |
|---|
Received for publication August 4, 2000.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  J. D. Jabro, A. D. Jabro, and R. H. Fox Accuracy and Performance of Three Water Quality Models for Simulating Nitrate Nitrogen Losses under Corn. J. Environ. Qual., July 1, 2006; 35(4): 1227 - 1236. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Agronomy Journal | Crop Science | |||
| Journal of Natural Resources and Life Sciences Education |
Vadose Zone Journal | ||||
| Journal of Plant Registrations | Journal of Environmental Quality |
The Plant Genome | |||
