Calibration
This chapter describes the calibration of WaTEM/SEDEM. In an efficient and semi-automated way.
Theoretical background
WaTEM/SEDEM can be calibrated for a specific study area by comparing the calculated sediment fluxes to the rivers with the observed sediment fluxes in a number of catchments. For every catchment in the calibration dataset an observed/measured sedimentflux must be present. In the model, only the ktc parameters can be choosen freely, and thus be used as calibration parameter. The ktc values only affect the transport capacity (TC) calculated by the WaTEM/SEDEM model. Therefor, this parameter can be adapted in order to calibrate the model.
In the model runs, two ktc values are used, namely: ktc-low and ktc-high. The first value, ktc-low, is used for land covers with low erosion potential (i.e. forest, pasture and grass strips), the latter, ktc-high, is used for land covers with high erosion potential (i.e. agricultural fields). Land covers with no erosion potential (i.e. infrastructure, rivers and open water) are automatically appointed with a very high ktc value (i.e. 9999).
In order to select the correct ktc values for a specific study area, WaTEM/SEDEM must be ran for a range of ktc values for all measurement areas in the dataset. The optimal combination of both ktc values is obtained by evaluating three criteria. This evaluation is done by the modeler in an external analysis and is not in the scope of this documentation.
The first criterium in the selection process is the calculation of the model efficiency \(ME\), defined by Nash and Sutcliffe (1970) as:
with
\(E_{obs,i}\): the observed sediment export for measurement point \(i\)
\(E_{sim,i}\): the simulated sediment export for measurement point \(i\)
\(E_{avg}\): the average observed sediment export of all \(n\) measurement points
The observed sediment export is typically determined by field measurements. In these field measurements, the total sediment load for an event and/or a total year is determined. It is important to note that the observations from the field measurements should be processed (by the user) in a way they are representative for the sediment load simulated by WaTEM/SEDEM. The simulated sediment load is the simulated load at a pixel (cfr. SediExport-raster) overlapping with the location of field measurements.
Model efficiencies can vary between \(-\infty\) and 1. An \(ME\) value smaller than zero means that the model is not efficient, i.e. the model delivers a result that is less accurate than the mean value of the observed values. An \(ME\) value of 1 can be interpreted as perfect model.
All calculated \(ME\) values for the different ktc-combinations can be visualised in a plot as shown by Van Rompaey et al. (2001).
Model efficiencies for different combinations of ktc-low and ktc-high (Van Rompaey et al., 2001)
Model simulations with a combination of ktc values with a high \(ME\) value are then further analysed. In the second criterium, the slope of the linear regression (with intercept 0) between the observed and simulated values is being analyzed. The simulated sediment export is considered ‘good’ if the calculated slope lies between 0.95 and 1.05. If not, the bias between model result and observation is considered too high (i.e. systematically more than 5% too high or too low).
It should be noted, however, that in areas with a predominant land cover (e.g. dominant cropland), only one ktc-value if often calibrated with a high ME, whilst the ME vs KTC-graph for the other value is rather flat (see also image above). In that case, a third and last criterium can be used, which examines the ratio between ktc-low and ktc-high. Multiple calibrations using different input data for river catchments in Flanders, Belgium, showed that this ratio typically lies between 0.25 and 0.35. Verstraeten et al. (2006) showed that when a ratio in this range is used, the simulated effect of grass buffer strips is regarded to be equal to the measured effectiveness of this type of erosion control measure, thus showing that the model is capable of predicting the impact of this type of soil and water conservation measure whereby differences in vegetation cover are crucial. However, in other environments, where the contrast in erosion potential between vegetated and non-vegetated land use classes is different compared to that between cropland and grassland/forest in a temperate region like Belgium, a different ratio between ktc-high and ktc-low can be sought for. Local insights into the different erosion potential may help in this respect.
Practical execution
WaTEM/SEDEM has a built-in calibration tool. This tool is an extenstion develloped during the work of Deproost et al. (2018). To use this tool, the user has to create a set of input rasters for every catchment in the calibration dataset and has to define all the options that are needed for the calibration and the future model runs. In the ini-file for every catchment that should be calibrated, the user has to enable the calibration option and define the range of ktc values.
The model will then loop over all combinations of ktc values in the defined range. First, a ktc map is created by the model for every ktc combination. Next, the full WaTEM/SEDEM model is run for all these combinations, for all the given catchments. Finally, a calibration file with the amount of sediment at each outlet of the model, for each combination of ktc values in the defined range is available for every catchment. These files can be processed by the user, through e.g. a python script, to calculate the \(ME\) and the other criteria, mentioned above, in order to select the best set of ktc-values for the study area.
References
Deproost, P., Renders, D., Van de Wauw, J., Van Ransbeeck, N., Verstraeten, G., 2018, Herkalibratie van WaTEM/SEDEM met het DHMV-II als hoogtemodel: eindrapport. Brussel. (in Dutch) https://archief.onderzoek.omgeving.vlaanderen.be/Onderzoek-1812384
Nash, J. E.; Sutcliffe, J. V. (1970). “River flow forecasting through conceptual models part I — A discussion of principles”. Journal of Hydrology. 10 (3): 282–290. https://doi.org/10.1016/0022-1694(70)90255-6
Verstraeten, G., Poesen, J., Gillijns, K., & Govers, G. (2006). The use of riparian vegetated filter strips to reduce river sediment loads: an overestimated control measure?. Hydrological Processes: An International Journal, 20(20), 4259-4267. https://doi.org/10.1002/hyp.6155