| Front page | | Contents | | Previous | | Next |

Calibration of Models Describing Pesticide Fate and Transport in Lillebæk and Odder Bæk Catchment

7 Discussion

7.1 Choice of model

The overall approach chosen has been to model catchments rather than single fields. This approach has resulted in a very thorough quality control of existing data from the catchments. This check had not been possible if only a part of the hydrological processes had been simulated. It is a point in favour of the model that it does not allow inconsistencies in input and measured data.

An important point for generation of water to the stream is the dynamics of the groundwater level. In case the project had chosen to limit the model to root zone models only it would have been difficult to generate the correct groundwater levels required as lower boundary condition for each of the soil columns used in the simulation.

The transport of "old" pesticides during low-flow due to transport with groundwater is actually captured by the model and could not have been with a de-coupled approach.

The limitation of the overall approach is the runtime of the models. Simulations take a long time and this has limited the amount of tests possible and the storage space required for intermediate files have limited the precision to less than that was the original goal.

7.2 Process descriptions

It is obvious, however, from the simulations that some process descriptions require more attention.

It is a critical issue that the colloid description implemented does not describe the experimental data collected by sub-project 4. The process description seriously underestimates the transport of strongly sorbing compounds. In order to be able to believe the model results for this group of compounds, a better validated process description will have to be implemented.

The model shows a tendency of build-up of the applied compound in the upper layer of the saturated zone. At the same time it often undervalues the peak concentrations of pesticides. Pesticide may thus arrive in the stream in concentration peaks that are too low but occur too long.

This has to do with the basic structure of the MIKE SHE-model. The unsaturated zone is divided into small calculation layers but the upper layer of the saturated zone has to be able to absorb the variation in the groundwater level over the year. This variation may be several meters (at least 2 m in Odder Bæk and 5 m in Lillebæk. see Figure 4.5. Figure 4.6. Figure 4.7. Figure 4.8 and Figure 4.11). Drain flow is generated from the saturated zone when the groundwater rises above drain level. The following example may describe the effect: A saturated layer of 2 m with 0.3 as porosity contains 600 mm water under drain level. 20 mm of water percolates to the saturated zone with a concentration of 20 μg/l. When this water is mixed with the water in the calculation layer, the concentration is reduced to 0.645 μg/l. Only 3.2% of this water then leave immediately in the drain. The rest of the water will move towards the stream or the catchment boundary over time. The concentration in the groundwater will therefore tend to build up.

Particularly in Lillebæk where the groundwater variations are large, this is considered a serious problem. However, it is also observed on the Lillebæk soils that the drains are placed in a rather impermeable layer, meaning that water will tend to saturate layers in the soil above the true groundwater level. Presently, MIKE SHE does not allow this water to move to the drains, as drains are not implemented in the unsaturated zone. It is expected that the modelled peaks would improve if drains would be implemented in the unsaturated zone. This is so because the mixing to a large extend would be avoided and because the transport time to the drain would be reduced. A secondary effect of this would be that the solute that leaves the soil column about the groundwater does not contribute to build-up of pesticide in the saturated zone later on.

It was already from the beginning of the project decided to compare a clayey catchment with a sandy catchment. The changes that have taken place during the project concerning the interpretation of the Odder Bæk catchment have violated some of the initial assumptions. The most serious assumption from a calibration point of view is that macropores were not included (by definition). However, there are indications of presence of macropores in Odder Bæk. substantiated by the fact that subproject 4 measured colloid transport in the soils. It is likely that the process is more relevant on the JB3-4- topsoils than on the JB1-2-topsoils, but it may also occur where a sandy topsoil lies on top of a more clayey subsoil.

As discussed for Lillebæk some of the observed pesticide peaks may not be caused by ordinary field applications. For the monitored drains, the connection between application and detection is never instantaneous – usually spring applications are found in the following autumn. The very fast responses observed during the spraying season may thus, at least in some cases, reach the stream through other pathways. Indications of point sources are found in other studies.

In the county of Bornholm, 10 agricultural enterprises were investigated. Water samples were taken in secondary groundwater close to the soil surface. Samples were taken in the area of the farm itself and not on the fields. Pesticide concentrations measured were between 3 and 1720 μg/l for pesticides in total. For single compounds the concentrations were lower (maximum 800 μg/l). The large concentrations assumed to be caused by a considerable use of pesticides, inappropriate washing places for spraying equipment or inappropriate handling of pesticides (Bay. 2001).

For enterprises spraying for farmers (maskinstationer. 42 investigations on 40 sites) the picture is similar (Amternes Videncenter 2002). Pesticides were detected on 94% of the sites investigated. It is thought that there are more point sources with large concentrations than the study indicated.

J. Kreuger (1998) finds that Atrazin. Hexazinon, Propyzamid, Simazin og Terbutylazin (and to some extent Bentazon og Cyanazin) stems from applications on non-agricultural land such as farmyards. In a single case a farmyard resulted in a concentration in the stream of 100 μg/l. In total more than 6 kg of Terbutylazin was washed out over 7 months. Large concentrations of a number of pesticides were collected in water from farmyards. The leaching continued over several months. Furthermore, two cases of spill in connection with filling or cleaning of equipment were identified. In the conclusion of the paper is written: Indeed. a substantial contribution of pesticide loss to stream water was from the application of pesticides in farmyards.

German results (Müller et al., 2000) have shown that a very large part of the pesticide application (>77%) arrived to the stream through the sewer as farm runoff was connected to the sewer system. In the final empirical model developed, only the applied amount of pesticide is a significant variable equal to 43 g a.i. for each field sprayer in the catchment. Measured concentrations were between 0 and 23.18 μg/l.

In the present study, one observation of bentazon in Odder Bæk could be a point source (2.9 μg/l). It is not related to a particular spraying and the level is considerably different from the rest of the measurements. For Lillebæk, the high occurrence of Terbutylazin in May 1999 is judged to be a point source. It is not possible to model the fast breakthrough observed in spite of the exaggerated macropores. Considering that only two small areas are sprayed upstream of the measuring station the thinning factor is likely to be substantial and the concentration just under the root zone should be in the order of mg/l. A high concentration of Diuron was measured simultaneously, - Diuron was sprayed on one of the two fields (and in a smaller dose). The concentration below the root zone should therefore be even higher to create the observed concentration at the measuring station. Furthermore, Simazin was observed in a high concentration without any recorded use of the pesticide.

On this basis it seems unrealistic to expect that no point sources are present in the material from the two catchments. The question is whether the model has been calibrated unrealistically to catch events, which are point sources rather than field applications.

7.2.1 Parameter choice

With respect to parameter choice for the compounds, very few modifications were done. Basically the values deducted for the compounds came from the PATE database and were estimated as recommended in FOCUS (2000). The only modification that was made was that the organic content in the Odder Bæk soils were halved compared to measurements. This sounds as a drastic change, and is so with certain limitations. First of all the area around Odder Bæk is organic soil, and there were no local measurements of its actual properties, the depth of the organic layer etc. The values here were therefore rather arbitrary from the beginning but they greatly influence the simulated pesticide transport. The simulation of isoproturon on the area of drain 51 is underestimating the actual load. It is however not very clear whether this is due to the organic content or because the estimated drain area for drain 51 is erroneous. Just around the delineated drain area are several fields, which receive the same pesticide.

The organic content of the soils from the soil profiles of Odder Bæk shows 4-5% organic matter. This is considered to be a very high content, particularly on a sandy soil used for arable farming. In connection with nitrate simulations, it has been discussed whether the high content is due to a lower reactivity of the organic matter. On the other hand, the change in organic content could also be a compensation for the fact that macropores were not included in the model.

7.3 Steering parameters

With respect to discretisation in space, the main issue is the size of the upper saturated zone-layer already discussed above. Of course the horizontal discretisation of 50m also introduces errors but nothing in the simulations has raised concern about the effect of the presently used grid size.

For Odder Bæk the discretisation in time was thought not to be critical because macropores are not included. All pesticide has to enter the drains through the unsaturated zone, and this process is relatively slow compared to the calculation timesteps used.

Discretisation in time is critical for macropore simulations. For this reason the water simulation file for pesticide simulations in Lillebæk was generated with a rainfall file that was "faked": it contained the same monthly rainfall as measured by the local station but it was divided into intensities (6 min) according to measurements on a different station. This method allowed generation of colloids and transport in macropores in a very detailed manner. However, due to the limitation with respect to storage, the information is averaged over the storage time step (16 hours for the unsaturated zone in Lillebæk). The mass is thus conserved but the peaks become too flat.

This is difficult to counteract as it requires a more detailed storage of information in the intermediate files. Presently, it has not been possible to break the 4.2 gigabyte limit of one file. There are solutions to this problem but it requires e.g. that the MIKE SHE code is changed so that it switches to a new file when the limit is approached.

 

| Front page | | Contents | | Previous | | Next | | Top |



Version 1.0 November 2004, © Danish Environmental Protection Agency