Agronomy Journal Journal of Natural Resources and Life Sciences Education
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

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 (2)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Bonesmo, H.
Right arrow Articles by Tremblay, G. F.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Bonesmo, H.
Right arrow Articles by Tremblay, G. F.
Agricola
Right arrow Articles by Bonesmo, H.
Right arrow Articles by Tremblay, G. F.
Related Collections
Right arrow Crop Growth and Development
Published in Agron. J. 97:32-40 (2005).
© American Society of Agronomy
677 S. Segoe Rd., Madison, WI 53711 USA

Agronomic Modeling

Timothy Yield and Nutritive Value by the CATIMO Model

III. Validation for Eastern Canada

Helge Bonesmoa, Gilles Bélangerb,*, Ed Charmleyc, Raynald Drapeaud, Dave B. McKenziee, Réal Michaudb and Gaëtan F. Tremblayb

a Norwegian Crop Res. Inst., Kvithamar Res. Cent., NO-7500 Stjordal, Norway
b Soils and Crops Res. and Dev. Cent., Agric. and Agri-Food Canada, 2560 Hochelaga Boulevard, Sainte-Foy, QC, Canada G1V 2J3
c Crop and Livestock Res. Cent., Agric. and Agri-Food Canada, Nappan, NS, Canada B0L 1C0
d Soils and Crops Res. and Dev. Cent., Agric. and Agri-Food Canada, 1468 Saint-Cyrille St., Normandin, QC, Canada G8M 4K3
e Atlantic Cool Climate Crop Res. Cent., Agric. and Agri-Food Canada, 308 Brookfield Road, St. John's, NF, Canada A1E 5Y7

* Corresponding author (belangergf{at}agr.gc.ca)

Received for publication July 2, 2003.

   ABSTRACT
TOP
 NOTES
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 REFERENCES
 
The Canadian Timothy Model (CATIMO) is a mechanistic simulation model of timothy (Phleum pratense L.) growth and nutritive value that features radiation interception and use efficiency, leaf and stem growth, leaf senescence, a N function based on the critical N concentration of whole plants, and cell wall (CW) concentration and digestibility of leaves and stems; the model was calibrated with measurements of timothy primary growth from one location (Fredericton, NB, Canada). In this paper, we compare the predictions of the CATIMO model with measurements from a total of six experiments conducted at four locations that cover the pedo-climatic conditions of eastern Canada (Lévis, QC; Nappan, NS; Normandin, QC; St. John's, NF). Across experiments, the root mean square errors (RMSE) for dry matter (DM) yield varied from 30.9 to 265.9 g DM m–2 and averaged 123.9 g DM m–2. Omitting one experiment in 1999, however, reduced the overall RMSE to 72.6 g DM m–2, which is very close to the 65.9 g DM m–2 reported for calibration. The RMSE values across experiments for N concentration (0.013 g g–1 DM), in vitro true dry matter digestibility (0.052 g g–1 DM), CW concentration (0.104 g g–1 DM), and CW digestibility (0.064 g g–1 CW) were higher than those obtained at the calibration site. At three of four locations, however, RMSE values were close to those of the calibration site. The CATIMO model seems robust enough to apply to situations different than those used for calibration, at least within eastern Canada.

Abbreviations: CW, cell wall • DM, dry matter • IVTD, in vitro true dry matter digestibility • LAI, leaf area index of green leaves • LWR, leaf to weight ratio • Nmmax, maximum nitrogen concentration • PAR, photosynthetically active radiation • RAW, readily plant available water • RMSE, root mean square error of estimation • RUE, radiation use efficiency • WAW, weakly plant available water


   INTRODUCTION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 REFERENCES
 
THE CANADIAN TIMOTHY MODEL,(CATIMO) is a mechanistic simulation model of timothy growth and nutritive value calculated on a daily basis, taking into account solar radiation, temperature, soil water, and N availability (Bonesmo and Bélanger, 2002a, 2002b). The growth module of CATIMO divides forage DM yield into leaves, stems, and senescent material. Separate relationships of CW deposition and decrease in CW digestibility are used for leaves and stems. Consequently, the model integrates the effect of the leaf to weight ratio (LWR) on forage nutritive value. The CATIMO model is one of few models that simulates both forage growth and nutritive value with linkages to plant components.

Process-based models like CATIMO are useful for understanding the complex interactions among plant growth, plant nutritive value, and environmental conditions. Furthermore, these models are also used in evaluating breeding strategies (Spitters and Schapendonk, 1990), N management (Gustavsson and Martinsson, 2001), and predicting crop growth and development in future warmer climates (Thornley and Cannell, 1997; Höglind et al., 2001). They can also be integrated in whole-farm simulators (Rotz et al., 1989) and agricultural decision support systems (Woodruff, 1992). The utilization of the CATIMO model for research, extension, or on-farm application depends on its ability to predict accurately timothy growth and nutritive value across a region broader than that in which it was developed and calibrated. The CATIMO model was calibrated with measurements taken weekly on timothy primary growth in four different years at one location (Fredericton, NB, Canada; Bonesmo and Bélanger, 2002a, 2002b). In this paper, we compare the predictions of the CATIMO model with measurements from a total of six experiments conducted at four other locations in eastern Canada.


   MATERIALS AND METHODS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS AND DISCUSSION
 REFERENCES
 
Field Experiments
Field experiments were conducted on timothy (cv. Champ or cv. Salvo) primary growth following winter at Lévis (QC), Nappan (NS), St. John's (NF), and Normandin (QC) (Table 1). Data from three or four N rates were extracted from the experiments at Lévis, St. John's, and Normandin, where N rates were replicated four times and the plot size was at least 4 m2. At Nappan, only one N rate was used. Crops were fertilized with N in early May at all sites. At Lévis, St. John's, and Normandin, a strip of at least 80% of the plot size was harvested weekly to determine DM yield; different plots were used each week. At Nappan, four strips (1.25 by 8 m) were taken weekly in each of three plots of 0.5 ha in size, and weekly measurements of LWR were also taken by separating leaves and stems from a subsample. At all sites, samples of approximately 500 g were taken for determinations of DM concentration. These samples were then used for the determinations of forage N concentration by the sulfuric acid (H2SO4)–hydrogen peroxide (H2O2) digestion method (Richards, 1993). Cell wall (CW) concentration, CW digestibility, and in vitro true dry matter digestibility (IVTD) were also determined. The IVTD was measured using the method based on rumen fluid digestion, followed by a CW determination of the postdigestion residues as described by Van Soest et al. (1966). The CW determinations of pre- and postdigestion samples were done by neutral detergent analysis as described by Van Soest et al. (1991). The CW digestibility was calculated from pre- and postdigestion CW dry weights.


View this table:
[in this window]
[in a new window]
  		Table 1. Initial conditions at the six field experiments used for validation.

 
The sampling at Lévis in 1996 was done on an experiment that started in 1995 when three rates of N fertilization were applied. Consequently, the initial soil inorganic N content to a 45-cm depth before the application of the three N rates in the spring of 1996 was 2.6 g N m–2 when no N was applied, 5.3 g N m–2 with an application of 4 g N m–2, and 8.3 g N m–2 with an application of 18 g N m–2 the previous year. At Lévis in 1999, the initial soil inorganic N content in spring was 6.9 g N m–2. At the other sites and years, the initial soil inorganic N content in spring was not measured, and we assumed a value of 5 g N m–2.

Water in soil layers to a 60-cm depth stored in the tension range 0.01 to 0.1 MPa was regarded as readily plant available (RAW), and water of the high tension storage from 0.1 to 1.5 MPa was assumed to be weakly plant available (WAW) (Table 1). Estimation of RAW was based on soil moisture retention curves obtained by the pressure plate technique (Richard, 1938, 1941). The WAW was also based on soil moisture retention curves at Nappan but was assumed to be 4.2 x RAW for clay soils and 3.4 x RAW for loam soils at the other sites, after data from Riley (1996). At the onset of primary growth, the soil was assumed to be at field capacity because of water from snowmelt and very low potential evapotranspiration. The assessments of growth start were based on field observations (Table 1).

Daily maximum and minimum air temperature, precipitation, and hours of bright sunshine were obtained from automatic climatic stations coordinated by Environment Canada at or near each experimental site. Daily global radiation was measured at St. John's in 1998 and Lévis and Normandin in 1999. At Nappan in 1997 and 1998, and Lévis in 1996, daily global radiation was estimated on the basis of the solar constant corrected for the effect of eccentricity, the integral of sine of solar height and atmospheric transmissivity calculated from the degree of cloudiness (Jones, 1992). Long-term averages of precipitation and mean daily temperature indicate a low risk of water deficiency to plants at all sites (Table 2). Among the four sites used for model validation, the temperature average for April to October was about 2°C lower at Normandin and St. John's than at Lévis and Nappan. The temperature and precipitation averages for April to October at Nappan were similar to those at the site used for model calibration (Fredericton). The average precipitation at Lévis and St. John's were greater than at the other sites, including the site used for calibration.


View this table:
[in this window]
[in a new window]
  		Table 2. Climate normals{dagger} for April through October at the four locations used for validation and at the site used for calibration.

 
Model Description
The model applies to the primary growth of timothy, which starts shortly after snowmelt and ends approximately 450°C-d (5°C basis) later when timothy is normally harvested at the early heading stage. The model has a daily time step. It includes modules for radiation use efficiency (RUE), leaf area growth and senescence, crop growth rate and DM production of leaves and stems, water, N, CW content, CW digestibility, and forage digestibility. Each module is briefly described below. A simplified diagram of the major components and their interrelationships is shown in Fig. 1, and the model is described in detail in Bonesmo and Bélanger (2002a)(2002b).



View larger version (63K):
[in this window]
[in a new window]
  		Fig. 1. Simplified diagram of the CATIMO model for the simulation of growth and nutritive value of timothy. T, temperature; PAR, photosynthetically active radiation; PET, potential evapotranspiration; RUE, radiation use efficiency.

 
Radiation Use Efficiency
The model is designed to simulate timothy growth by assuming a constant potential RUE [g DM MJ–1 photosynthetically active radiation (PAR)]. The actual RUE is computed daily using functions that decrease RUE due to suboptimal growing conditions of daily PAR and temperature and deficiencies of N and water.

Leaf Area Growth and Senescence
The leaf area index of green leaves (LAI) is obtained by integrating the daily net result of the leaf area growth rate and senescence rate (Spitters and Schapendonk, 1990). In the early stage, leaf area increases exponentially with temperature. Above a critical LAI when new leaves do not increase light interception, the phase of exponential leaf area growth changes into a linear growth phase (Goudriaan and Monteith, 1990). In the linear phase, the model calculates leaf area growth using the product of the increase in leaf weight (g leaf DM m–2 d–l) and specific leaf area of new leaves (m2 leaf g–l leaf DM). The specific leaf area of new leaves is calculated as a function of temperature based on the assumption that it increases up to an optimum temperature and then decreases (Solhaug, 1991). Leaf senescence is assumed to begin when the temperature sum above a base temperature of 0°C, from initial growth, reaches 260°C-d (Bélanger, 1996), after which it is assumed to be affected by temperature.

Crop Growth Rate and Dry Matter Production of Leaves and Stems
The daily crop growth rate (g DM m–2 d–1) is calculated by multiplying the actual RUE by the amount of PAR intercepted. The daily values of intercepted PAR are derived by assuming that light interception increases with LAI according to a negative exponential function, with an extinction coefficient. Dry weights (g DM m–2) of the green leaves, senescent leaves, and stems are obtained by integrating their respective growth and death rates. The proportion of the daily crop growth rate partitioned to leaves is related to the phenological development, described by the temperature sum. The senescence rate of leaves in terms of dry weight is zero for temperature sums lower than 260°C-d. For temperature sums higher than 260°C-d, the senescence rate is defined using the same relative senescence rate that applies to LAI, multiplied by the dry weight of the green leaves. The disappearance rate of senescent leaves is assumed to be similar to the senescence rate, but with a time delay. The time delay is required to exclude senesced leaves from LAI but to keep them on the plants until they fall off for the calculation of forage nutritive value.

Water
Drought stress is calculated as the ratio between actual evapotranspiration (Ea) and potential evapotranspiration (Ep) from plants, where Ep (mm d–l) and Ea (mm d–l) are derived from a model that estimates soil water evaporation and plant evapotranspiration separately (Ritchie, 1972); the model is expanded to include a soil water budget (Skjelvåg, 1981). The combined model calculates potential and actual evapotranspiration from plants on the basis of potential evapotranspiration, LAI, and the content of RAW and WAW in the root zone.

Nitrogen
In CATIMO, N deficiency affects the parameters associated with RUE and leaf area expansion, including relative growth rate of leaves and specific leaf area. The crop N status is expressed as the relative N concentration calculated as the ratio of actual N concentration to critical N concentration (Bélanger and Gastal, 2000). The effect of N status on RUE is calculated according to a function given by Bélanger and Richards (1997). The actual crop N concentration is calculated by assuming that the fraction of absorbed N remaining in the aboveground biomass increases with increasing N status of the crop (Brégard et al., 2000). The total N in aboveground and belowground biomass is obtained by integrating the daily rate of N absorption. The daily rate of N absorption is assumed to be at the minimum of crop demand and soil supply. The N demand (g N m–2), including aboveground and belowground biomass, is characterized by the difference between the amount of N required to go from limited conditions to the maximum limit of N absorption (Brisson et al., 1998). The potential supply of N to the crop (g N m–2) is assessed by the current soil mineral N content and an availability factor for soil N.

Cell Wall Concentration
The CW concentration of green leaves and stems is obtained by integrating the proportion of the respective daily growth rates partitioned to CW, the daily rates of conversion of cellular content into CW, and the daily death rate of leaves. Since temperature affects the rates of CW deposition (Fales, 1986), both the partitioning and the conversion in leaves and stems are assumed to be affected by temperature. The rates of CW deposition are assumed to decrease in both leaves and stems when the crop was N stressed (Bélanger and McQueen, 1999). A multiplicative index based on the relative N concentration is introduced to account for the effect of N stress on CW deposition.

Cell Wall Digestibility
The daily decrease in CW digestibility is indirectly related to increasing phenological development but also directly related to increasing temperature (Ford et al., 1979). Thus, the CW digestibility of leaves and stems is determined from an initial maximum value of CW digestibility (g g–l CW) and a daily rate of decrease (g g–l CW d–l) for leaves and for stems as related to the daily mean temperature. Early in the growth cycle, there are no true stems (Bélanger and Richards, 1995). Thus, daily rate of decrease for stems is assumed to be similar to daily rate of decrease for leaves when the temperature sum, from spring growth initiation, is below that required for the start of true stem elongation. As for CW deposition, the rates of decline in CW digestibility are assumed to decrease in both leaves and stems when the crop was N stressed, and the multiplicative index based on the relative N concentration accounts for the effect of N stress on the daily rate of decrease in CW digestibility.

Forage Digestibility
In CATIMO, the crop is considered to consist of green leaves, dead leaves, and stems including leaf sheaths. Green leaves and stems are characterized for CW concentration and CW digestibility. We assume a DM digestibility of 0.98 g g–l DM for the cellular content of green leaves and stems (Van Soest, 1982). Dead leaves are assumed to consist of CW only, with a DM digestibility of 0.70 g g–l DM (Duru, 1997). The DM digestibility of the forage is calculated by combining the DM digestibility of green leaves, dead leaves, and stems with their weight. The basic idea is the separation of leaf and stem DM into cellular content and CW constituents. The cellular content of leaves and stems is almost completely digestible, and although the CW constituents have limited digestibility, this does not inhibit the digestion of the cellular content (Deinum, 1973).

Model Performance
Performance of the CATIMO model was assessed using both a deviation-based statistic (RMSE) and a linear regression analysis between simulated and measured values to quantify possible over- or underestimation by the model.


   RESULTS AND DISCUSSION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
REFERENCES
 
The simulated DM yield showed good agreement with the measurements for the nonlimiting N conditions at Lévis in 1996, Nappan in 1997, St. John's in 1998, and Normandin in 1999 (Fig. 2; Table 3). A minimum-value approach was used for the growth-limiting factors (water and N) with the respective reduction functions f(w) and f(N). Under nonlimiting N conditions, water stress became the predominant reduction function. Under those conditions, the simulations were affected by water stress at Lévis in 1996, at Normandin in 1999, and at St. John's in 1998 with respective average f(w) values in June of 0.84, 0.73, and 0.90. At those sites, the calculated effect of the water stress on forage DM yield worked satisfactorily, with RMSE less than 16% of the average DM yield and a slope close to 1.0 (Table 3). The model, however, underestimated DM yield at the end of the growth period at Nappan in 1998 (Fig. 2); this might be primarily caused by an overestimation of the water stress with a low value of f(w) (0.76) in June.



View larger version (26K):
[in this window]
[in a new window]
  		Fig. 2. Measured and simulated cumulative forage dry matter (DM) yield of timothy grown under limiting N conditions (open symbols, broken lines) and nonlimiting N conditions (closed symbols, solid line) in six separate experiments. N0, N4, N6, N7.5, N12, and N18 represent 0, 4, 6, 7.5, 12, and 18 g N m–2, respectively.

 

View this table:
[in this window]
[in a new window]
  		Table 3. Summary statistics of key model attributes.

 
Under the most limiting N conditions at Lévis in 1996 and at St. John's in 1998, the growth was more limited by N than by water so that the water stress did not affect the simulations. Average values in June of the reduction function for suboptimal N [f(N)] when no N was applied were 0.53 at Lévis in 1996 and 0.69 at St. John's in 1998. The assumed initial soil inorganic N at St. John's in 1998, however, was probably too high, resulting in an overestimation of forage DM yield under limiting N conditions. The use of a lower value, 2 g N m–2, gave a better growth simulation; the RMSE for forage DM yield was reduced from 73.8 to 48.7 g m–2.

The large underestimation of the growth at Lévis in 1999, especially between the second and third week, cannot be fully explained by an overestimation of the growth reduction functions related to water stress [f(w)] or N deficiency [f(N)]. Failures with the growth-defining factors may rather be the most important cause. The global radiation (20.9 MJ m–2 d–1) and the temperature (18.5°C) conditions in June at Lévis in 1999 did not differ much from those at Lévis in 1996 (20.5 MJ m–2 d–1, 17.8°C). The global radiation in June at Lévis, however, was higher than 20 MJ m–2 d–1 for 21 d in 1999 but only 17 d in June 1996. Hence, we speculate that it was primarily the reduction of RUE at high daily radiation, f(PAR), that was decreasing the simulated growth at Lévis in 1999; the average f(PAR) in June was 0.75. Leaving f(PAR) out gave a much better simulation result for Lévis in 1999, decreasing the RMSE from 265.9 to 85.4 g DM m–2. In our model, air temperature was used. Under high radiation at Lévis in 1999, however, plant canopy temperature may have been higher than air temperature. This could also have contributed to the underestimation of DM yield at this site.

Across experiments, the RMSE varied from 30.9 to 265.9 g DM m–2, with an average of 123.9 g DM m–2 (Table 3). This is about twice the RMSE reported for the calibration site (Bonesmo and Bélanger, 2002a). However, omitting Lévis in 1999 reduced the overall RMSE to 72.6 g DM m–2, which is closer to the 65.9 g DM m–2 reported for the calibration.

The relationship between simulated and measured N concentrations is similar to that obtained with the model calibration (Bonesmo and Bélanger, 2002a), that is, an overestimation of N concentration under nonlimiting N conditions and an underestimation of N concentration under limiting N conditions (Fig. 3; Table 3). The over-estimation is particularly important for St. John's in 1998 and Normandin in 1999. There was no yield response above 12 g N m–2 at St. John's and no yield response to N fertilization at Normandin. In situations of no yield response, nitrates can accumulate, and our model seems to overestimate this accumulation of nitrates. It is most likely that the model overestimated the N demand because of the maximum N concentration (Nmmax), which might be too high at 0.071 g N g–1 DM. In wheat (Triticum aestivum L.) simulation models using the concept of critical N concentration, Nmmax is close to 0.050 g N g–1 DM (Hansen et al., 1991; Porter, 1993; Stockle and Nelson, 1996). Using a Nmmax of 0.050 g N g–1 DM reduced the RMSE from 0.021 to 0.008 g N m–2 at St. John's in 1998 and from 0.013 to 0.006 g N m–2 at Normandin in 1999.



View larger version (19K):
[in this window]
[in a new window]
  		Fig. 3. Simulated N concentration plotted against measured values for timothy grown under limiting (open symbols) and nonlimiting (closed symbols) N conditions in six separate experiments. DM, dry matter.

 
The simulation of LWR at Nappan in 1997 and 1998 was satisfactory (Fig. 4), which increases confidence in the sound biological basis of the model. A correct simulation of LWR is crucial for the prediction of forage digestibility and CW concentration. The trends in the relationship between simulated and measured IVTD and CW concentration were similar to those obtained with the model calibration, that is, an overestimation of the CW concentration, particularly at Lévis in 1996 and 1999 (Fig. 5; Table 3). The CW digestibility, however, was underestimated at Lévis in 1996 and 1999, whereas it tended to be overestimated at Nappan and St. John's (Fig. 5). The overestimation of CW concentration and the underestimation of CW digestibility at Lévis resulted in an underestimation of IVTD, especially at end of the growth period. The overestimation of CW concentration could be attributed to an overestimation of the proportion of stems and/or an overestimation of CW deposition under some conditions. It could be that the model partitions too much of the DM growth to the stems, which have a higher CW concentration than the leaves. However, the simulation results from the calibration site, where the stem DM was measured, indicated that the CATIMO model tended to underestimate the stem growth at end of primary growth cycle (Bonesmo and Bélanger, 2002a). Thus, a more likely explanation is that the model allows a too high CW deposition at high temperatures. Both the daily growth partitioned to CW of leaves (GCWL) and stems (GCWS) and the daily conversion rate of cellular content into CW for leaves (RCCL) and stems (RCCS) increase with increasing temperature. The daily temperatures during the growth period at Lévis (average temperatures: 1996, 14.1°C; 1999, 16.6°C) were mostly higher than those at the other sites (average temperature 12.5°C) and at the model calibration site (average temperature 13.6°C). The same explanation is also applicable for the underestimation of the CW digestibility at Lévis. Similar to the increase in CW deposition, the decline in CW digestibility increases with increasing temperature in the CATIMO model. This research represents one of the first efforts in modeling CW concentration and CW digestibility in forage grasses. Further research is necessary to obtain more accurate quantitative relationships between climate conditions and CW deposition and changes in CW digestibility.



View larger version (15K):
[in this window]
[in a new window]
  		Fig. 4. Measured (symbols) and simulated (lines) leaf weight ratio (LWR) of timothy grown at Nappan in 1997 and 1998. The parameter "a" represents intercept, and the parameter "b" represents the slope of the linear regression between simulated and measured values. RMSE, root mean square error of the prediction; RMSEn, normalized RMSE, calculated as RMSE divided by the average value of the attribute.

 


View larger version (21K):
[in this window]
[in a new window]
  		Fig. 5. Simulated forage in vitro true dry matter digestibility (IVTD), cell wall (CW) digestibility, and CW concentration plotted against measured values for timothy grown under nonlimiting and limiting N conditions in six separate experiments. DM, dry matter.

 
The RMSE values obtained with this independent data set were mostly higher than those obtained with the calibration data. The overall RMSE values for DM yield, N concentration, CW concentration, and CW digestibility of the test sites were approximately twice (Table 3) those obtained at the calibration site (Bonesmo and Bélanger, 2002a, 2002b). For some experiments, however, a lower or similar RMSE was obtained. For DM yield under nonlimiting conditions, the RMSEs at Lévis in 1996 (77.3 g DM m–2), at Nappan in 1997 (86.3 g DM m–2), at St. John's in 1998 (30.9 g DM m–2), and at Normandin in 1999 (53.1 g DM m–2) were close to that obtained at the calibration site (65.9 g DM m–2). For N concentration, the RMSEs at Nappan in 1998 (0.003 g N g–1 DM), at Lévis in 1996 (0.004 g N g–1 DM) and 1999 (0.003 g N g–1 DM) under limiting N conditions, and at St. John's in 1998 (0.006 g N g–1 DM) under limiting N conditions were less than that obtained at the calibration site (0.009 g N g–1 DM). The RMSEs for IVTD at Normandin in 1999 (0.014 g g–1 DM), for CW concentration at Nappan in 1997 (0.021 g g–1 DM) and 1998 (0.039 g g–1 DM), and at St. John's in 1998 under nonlimiting (0.042 g g–1 DM) N conditions were also less than those obtained at the calibration site.

The CATIMO model seems robust enough to apply to situations different than those used for the calibration, at least within eastern Canada. The validation with an independent data set, however, confirmed that the problems observed during the calibration are still present. The model was calibrated with data from 4 yr at one location, and information from the literature on timothy was used to assess the parameters. This approach for the calibration provided some very interesting and promising results, but some problems remain. Some of those problems are related to gaps of knowledge of the complex interaction among timothy growth, nutritive value, and environmental conditions.

The development and validation of the CATIMO model has been an effective approach to identify knowledge gaps and prioritize research goals. For example, we discussed possible errors related to CW deposition, Nmmax, and the calculation of the actual RUE. The RUE of grasses can be very high under low light conditions. Ryegrass (Lolium perenne L.) grown under low light conditions in glasshouse produced DM with an efficiency of 5.0 g MJ–1 PAR (Schapendonk et al., 1997), which is considerably higher than the potential RUE of 3.0 g MJ–1 PAR used in CATIMO. In the natural environment, however, plants average daily variations in irradiance over long time periods so that the small daily variations induced in RUE are not apparent (Sinclair and Muchow, 1999). Furthermore, in June, high values of PAR with close-to-optimum temperatures are observed, and this was particularly the case at Lévis in 1999; our f(PAR) reduction function, particularly the slope coefficient, PARcoeff, probably prevents the model from accounting for those specific conditions that occurred in 1999. This seems to be a deficiency of the model. The Nmmax has been studied in much less detail than the critical N concentration (Nc), which has received a great deal of attention in several crops (Jeuffroy et al., 2002). Our value of Nmmax, 0.070 g m–2, is similar to that used by Gustavsson et al. (1995) in timothy but considerably higher than those used in wheat models. A model evaluation of variations in single parameter values must, however, be followed by a recalibration of the model since plant growth is influenced by multiple parameters. Thus, to ensure the flexibility for modeling growth and nutritive value under contrasting conditions, further development of the model must include a wider basis for calibration, including the data used in the current validation.

A prospective use of the CATIMO model in breeding for a higher nutritive value of timothy could be to study whether it is the morphological component (amounts of leaves, stems, and senescent plant material) or the direct changes in the CW digestibility that causes the largest effect on forage digestibility. A model study with perennial ryegrass indicated that morphological characteristics had limited effects on grass digestibility; direct changes in the CW composition and digestibility might cause larger effects (Groot and Lantinga, 1999). If the model were to be used in whole-farm simulators, in agricultural decision support systems, or for the prediction of the impact of global warming on milk and forage production (Topp and Doyle, 1996), two main features must be added to the model. For primary growth, the growth start represents a problem. In the present model, the date of the growth start is an input. Furthermore, the model is calibrated and tested only on spring growth. With timothy, however, the summer growth under Canadian conditions can represent up to 40% of the seasonal yield (Bélanger and Richards, 1997). This highlights the need to develop our model for summer growth. For example, the maximum growth rate, and thus the potential RUE of timothy summer growth, declines with advance in stage of phenological development at first harvest (Bonesmo and Skjelvåg, 1999).

A more direct use of our model could be to include modules from CATIMO into other models. The CATIMO functions used for CW deposition and CW digestibility of leaves and stems could be combined with growth models that provide simulation of leaf and stem DM and dead leaves. Such combination was demonstrated by Höglind and Bonesmo (2002) using the LINGRA-timothy model developed for Scandinavian conditions (Höglind et al., 2001). Simulations with the combined model showed good agreements with measurements for timothy grown in Norway for CW concentration, whereas the decline in CW digestibility was underestimated, probably due to daylength effects.


   NOTES
TOP
 NOTES
ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 REFERENCES
 
Contrib. no. 773, Agric. and Agri-Food Canada.


   REFERENCES
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 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 (2)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Bonesmo, H.
Right arrow Articles by Tremblay, G. F.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Bonesmo, H.
Right arrow Articles by Tremblay, G. F.
Agricola
Right arrow Articles by Bonesmo, H.
Right arrow Articles by Tremblay, G. F.
Related Collections
Right arrow Crop Growth and Development

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Crop Science Vadose Zone Journal
Journal of Natural Resources
and Life Sciences Education		 Soil Science Society of America Journal
Journal of Plant Registrations Journal of
Environmental Quality		 The Plant Genome