CHAPTER 9



KNOWLEDGE GAPS









The nitrogen components of the model have been constructed with the philosophy of maintaining simplicity and a state of balance with the remainder of the model. The model itself has been constructed within the framework of keeping the number of inputs required to a minimum and to those which are easily attainable. This latter constraint has generally precluded the use of laboratory procedures to characterize soil and site specific rate constants for the nitrogen transformations. In some instances predictions would undoubtedly be improved if appropriate specific rate constants were used as inputs, but this would detract from the generality of the model. Methods of estimating these rate coefficients from other more commonly available soil properties are needed if improvement in the predicted rates of the various transformations is to be made.

The CERES model does not simulate losses of nitrogen via ammonia volatilization. While arguments can be made that under conditions of good fertilizer management (placement of fertilizer especially) these losses should be small, there are conditions where losses can become substantial. Ammonia loss from surface applications of urea to alcareous soil and from anhydrous ammonia bands can be large. Several models exist within the literature to simulate these losses (eg. Rachhpal-Singh and Nye,1986 and McInnes et al) but all operate at levels of detail beyond that used in CERES. A simple method of predicting ammonia loss that requires only modest inputs is sorely needed. Among the factors known to be important are soil moisture content, soil pH, soil cation exchange capacity, temperature and soil hydrogen and ammonium ion buffering capacities.

The procedures to describe leaching of nitrate within the model are very simplistic and as pointed out in some circumstances may be inadequate. Addiscott (1981) has described a model that partitions water and solutes between a mobile phase and a retained phase. The partitioning was made on the basis of a two bar moisture content. The CERES model does perform some partitioning of the moving water on the basis of the whole profile saturated water conductivity (SWCON) and also on the partitioning between water below the drained upper limit and above it. Appropriate procedures to enable the partitioning of nitrate between these various soil water phases are required. The procedures would need to account for differences in equilibration of solute between these phases which would occur in different soils.

One problem encountered with the nitrogen components of the model was the inability to accurately predict mineralization of nitrogen on certain unusual soils. The model under predicted the mineralization rate on virgin soils which have been recently cultivated. In other instances where "protected" organic matter occurs in some andosols the model over predicts mineralization. Currently provision is made to adjust the mineralization rate with an arbitrary user-supplied scalar DMOD for these circumstances. An appropriate way of identifying these circumstances or adjusting the mineralization rate constant (DMINR) for these soils is required. Laboratory procedures (eg Stanford et al 1972) exist for this, but the estimates are not commonly available. A surrogate value which could be estimated from other soil properties may be possible. Another problem which exists with the mineralization calculations is concerned with determining the appropriate way to initialize the fresh organic matter pools. These are currently initialized as 20% carbohydrate, 70% cellulose and 10% lignin.

When fertilizer is placed in a band or point placed, the model assumes uniform incorporation within the layer in which the fertilizer is placed. When high concentrations of fertilizer affect the chemistry of the soil in such a way as to substantially alter the nitrogen transformation rates this simplification will be in error. These fertilizer practices are of more concern in wide-row crops and should not greatly affect the simulation of nitrogen dynamics in wheat cropping systems.

One significant problem encountered when attempting to simulate the accumulation of nitrogen in the grain was how to account for genotypic differences in grain protein accumulation. The model in general terms can account for differences in temperature, soil water status, and plant N status on grain N accumulation. However, the simulations are often poor. Several researchers have pointed to cultivaral differences (Hunt 1984). A further genetic coefficient describing the rate of grain N accumulation per degree day or alternatively some factor defining cultivaral differences in the minimum vegetative N concentration during grain filling is probably required. Cultivaral differences also occur in the critical concentration growth stage relationships. However, other than differentiating between winter and spring wheat cultivars, no attempt has been made to separate these out.

In some experiments nitrogen has been reported to have an effect on growth duration of the crop. Often high nitrogen treatments are reported as having a delayed maturity. One possibility for this is that under a higher plane of nutrition tiller survival is greater. Since anthesis and maturity on higher order tillers are later than on the main stem, the mean anthesis and maturity date for the whole plant are delayed. An alternative hypothesis advanced by Davidson and Campbell (1982) is that the increased canopy growth associated a higher N rate in some way may contribute to a cooler meristem either through greater shading or via greater transpiration. Further research is needed to determine the mechanisms of this delay before it can be modelled.