Diffuse losses of phosphorus (P) from arable land are high, but since they are not
possible to measure at national scale they need to be estimated by use of nutrient
transport models. A new methodology for estimation of P losses from agriculture
has been developed for the purposes of HELCOM´s Pollution Load Compilation.
This methodology is based on P export coefficients which are calculated with
ICECREAMDB model. A matrix of export coefficients includes 22 leaching
regions, 15 crops, 10 soil types, 3 slope classes and 3 soil P content classes.
Considering limitations of input data and the validity of assumptions which were
made in the calculations, as well as the limited testing of the applied model for
Swedish conditions, an analysis of the discrepancies between calculated results and
measured values was performed in this project to identify potential weaknesses and
possible improvements to develop future calculations. Small catchments dominated
by agriculture were studied in order to focus on diffuse sources from agriculture.
Additionally, in order to minimize uncertainties connected to the P retention in
lakes, a selection of catchments without lakes was done. Besides this analysis of
discrepancies, additionally two issues were studied. The first issue is the seasonal
and/or flow related differentiation of agricultural export coefficients which was
done by statistical analyses of measured data. The other issue considers the
evaluation of calculated high P losses in leaching region 9 which was done by
comparison of modelled results and available measured values.
Three main explanations were found for discrepancies between measured and
modelled values:
1. Input data is not representative or its resolution is too low to properly
describe conditions in the certain catchment
2. Parametrisation of the ICECREAMDB model is inadequate and or the
assumptions that were made are too uncertain
3. The important processes in the ICECREAMDB model are not properly
described and should be changed to improve results
The problem with weak input data is outside the frames of this project but since
weak input data influence our possibilities to evaluate model performance we
suggest testing of ICECREAMDB model in catchments where we have better data
or such data is easily collected.
In order to minimize uncertainties connected to class definition regarding slope and
soil P content, use of regression equations is suggested since it allows use of
specific values for each given catchment.
5
Several possibilities for further model development are also identified in this
project. Some of the most important are improvement of the description of P
chemistry and distribution of P in different P pools. Also, model application might
be improved by better parametrisation of factors governing apportionment between
surface runoff and infiltration as well as generation of soil particles. Additionally, a
wider calibration and testing of ICECREAMDB and export coefficients are
suggested. In the first step, set-up, parametrisation, calibration and validation of
ICECREAMDB should be done on the field scale with the available water quality
monitoring data. Then, in the second step set-up, parametrisation, calibration and
validation of a source apportionment model for a number of small agricultural
catchments is suggested in order to test and validate calculated export coefficients
and to give feedback to field-scale modelling.
Performed statistical analyses show that there is little statistical support for
introduction of seasonal or flow-dependent export coefficients instead of the
current annual average.
A comparison between measured and modelled values in leaching region 9 shows a
reasonable agreement where the calculated values are underestimated rather than
overestimated.