The 0.05 degrees quasi-global (-180,180,90,-60) implementation of the LISFLOOD OS model was calibrated using in-situ discharge observations from stations with a minimum drainage area of 500 km2 and at least 4-years-long time series of measurements more recent than 01 January 1980.
The 1996 selected calibration stations entailed 47.5 % of the quasi-global domain (Figure 1, blue area). The parameter values of these catchments were identified using the Distributed Evolutionary Algorithm for Python (DEAP, Fortin et al. 20121). The parameter values of the catchments for which in situ discharge data were not available (Figure 1, yellow area) were estimated by parameter regionalisation. This combined calibration approach delivered 14 parameter maps with quasi-global extent. An overview of the calibration methodology and parameters used in CEMS-GloFAS is available here. This page provides specific information about GloFAS v4.
Figure 1 - In blue the area of the semi-global domain for which discharge observations were available; in yellow the area of the semi-global domain for which discharge observations were NOT available (a parameter regionalisation approach was used in these areas) for GloFAS v4 calibration.
Parameter estimation for catchments with discharge data
The Distributed Evolutionary Algorithm for Python (DEAP, Fortin et al. 20121) was used to explore the parameter space and identify the parameter set leading to the highest value of the modified Kling Gupta Efficiency (KGE', Gupta et al., 20092), as implemented by the open-source calibration tool.
When multiple calibration points were available in one basin, the calibration protocol followed a top-down approach from head-catchments to downstream catchments; each segmentation of the basin area is called inter-catchment. Figure 2 shows the fragmentation of the area with available discharge observations into inter-calibration catchments.
Figure 2 - Fragmentation of GloFAS semi-global domain into LISFLOOD calibration inter-catchments.
The size of the inter-catchments was mainly driven by data availability. The largest inter-catchment was located in the Congo basin, with a drained area larger than 2.500.000 km2. Figure 3 shows the distribution of the area of the calibration inter-catchments, the median value was 14.000 km2.
Figure 3– Calibration stations for GloFAS v4: distribution of the area of the LISFLOOD calibration inter-catchments. Note the logarithmic scale of the x-axis.
The time series considered for calibration covered 40 years, from 01/01/1980 to 31/12/2019. Each calibration station used a different calibration period, depending on length of available discharge observations. The minimum number daily observation data use in calibration was the equivalent of 4 years. A spin-up period of three year was added to each model run to ensure the correct initialization of the hydrological model. Nevertheless, some exceptions were implemented: GloFAS v4.0 calibration data - Copernicus Emergency Management Service - CEMS - ECMWF Confluence Wiki. If a period lower than 8 years was available, all the data were used for calibration. For periods between 8 and 16 years, the last 8 years were used for calibration. Finally, if more than 16 years were available, the discharge record was split in two. The most recent period was used for the calibration because the most recent forcing data and discharge observations are expected to have lower uncertainty and to provide a closer representation of the climatic and hydrological conditions of the forecast period. Figure 4 shows the distribution of the length of the time series used for calibration.
Figure 4 – Calibration stations for GloFAS v4: number of years used in calibration.
Calibration parameters
14 LISFLOOD parameters were simultaneously calibrated for each catchment, with the purpose to optimize the modelling of snow melt, water infiltration into the soil, surface water flow, groundwater flow, lakes and reservoirs dynamics. Feasible parameter ranges were defined for each parameter used in calibration to obtain more physically realistic calibrated parameters. Table 1 lists the calibration parameters, the acceptable range of values, and the default values.
Parameter name | Description | Symbol | Min | Max | Default |
SnowMeltCoef | Snow melt rate in degree day model equation [mm/(C day)] | M | 2.5 | 6.5 | 4 |
b_Xinanjiang | Exponent in Xinanjiang equation for infiltration capacity of the soil [-] | INFact | 0.01 | 5 | 0.5 |
PowerPrefFlow | Exponent in the empirical function describing the preferential flow (i.e. flow that bypasses the soil matrix and drains directly to the groundwater) [-] | Dpref,gw | 0.5 | 8 | 4 |
UpperZoneTimeConstant | Time constant for upper groundwater zone [days] | Qugw | 0.01 | 40 | 10 |
GwPercValue | Maximum percolation rate from upper to lower groundwater zone [mm/day] | Dugw,lgw | 0.01 | 2 | 0.8 |
LowerZoneTimeConstant | Time constant for lower groundwater zone [days] | Qlgw | 40 | 500 | 100 |
LZThreshold | Threshold to stop outflow from lower groundwater zone to the channel [mm] | Qlgw | 0 | 30 | 10 |
GwLoss | Maximum loss rate out of lower groundwater zone expressed as a fraction of lower zone outflow [−] | Qlgw | 0 | 1 | 0 |
QSplitMult | Multiplier to adjust discharge triggering floodplains flow [-] | Qch | 0 | 20 | 2 |
CalChanMan1 | Multiplier for channel Manning's coefficient n for riverbed [−] | Qch | 0.5 | 2 | 1 |
CalChanMan2 | Multiplier for channel Manning's coefficient n for floodplains [−] | Qch | 0.5 | 5 | 1 |
adjust_Normal_Flood | Multiplier to adjust reservoir normal filling (balance between lower and upper limit of reservoir filling). [-] | Qres | 0.01 | 0.99 | 0.8 |
ReservoirRnormqMult | Multiplier to adjust normal reservoir outflow [−] | Qres | 0.25 | 2 | 1 |
LakeMultiplier | Multiplier to adjust lake outflow [−] | Qlake | 0.5 | 2 | 1 |
Table 1 - LISFLOOD calibration parameters for GloFAS v4.0.
Parameter regionalisation
Catchments for which in situ discharge data was not available entailed 52.5% of the quasi-global domain (Figure 1, yellow area). For these catchments, a parameter regionalisation approach was implemented to estimate the parameters that control snow melt, infiltration, runoff, groundwater, and routing (11 parameters in total, see Table 1).
Calibrated parameter values were transferred from a calibrated catchment (the "donor") to each uncalibrated catchment. For each uncalibrated catchment, the donor catchment was identified according to a proximity criterion accounting for both climatic similarity and geographical proximity (Parajka et al 20053, Beck et al. 20164, see also CEMS-Flood hydrological model calibration method). Parameters controlling the local behaviour of lakes and reservoirs cannot be transferred from donor catchments to target catchments: default lakes and reservoirs parameter values (see Table 1) were used in uncalibrated catchments.
A leave-one-out cross validation approach for a subset of 739 gauged catchments was used to verify the performances of the regionalisation approach. Moreover, stations with observation time series shorter than 4 years (excluded from calibration) were used to verify the benefit of using the regionalised parameters over the default parameter values. As an example, Figure 5 allows to compare the observed discharge (black line) with the model results generated using the regionalised parameters (red line) and the default parameters (blue line) for one small catchment on the Black Sea (left) and one small catchment in Siberia (right).
Finally, it is here noted coastal and endorheic catchments with drainage area smaller than 500 km2 are modelled using the default parameter values (Table 1).
|
|
Figure 5 – Observed (black) and modelled discharge using regionalised parameters (red) and the default parameter set (blue) for two not calibrated catchments located on the Black Sea (left) and in Siberia (right)
References
1 Fortin, F. A., De Rainville, F. M., Gardner, M. A. G., Parizeau, M., & Gagné, C. (2012). DEAP: Evolutionary algorithms made easy. The Journal of Machine Learning Research, 13(1), 2171-2175. https://jmlr.org/papers/volume13/fortin12a/fortin12a.pdf
2 Gupta, H. V., Kling, H., Yilmaz, K. K., & Martinez, G. F. (2009). Decomposition of the mean squared error and NSE performance criteria: Implications for improving hydrological modelling. Journal of hydrology, 377(1-2), 80-91. https://www.sciencedirect.com/science/article/pii/S0022169409004843?via%3Dihub
3 Parajka, J., Merz, R., and Blöschl, G. (2005). A comparison of regionalisation methods for catchment model parameters, Hydrol. Earth Syst. Sci., 9, 157–171, https://doi.org/10.5194/hess-9-157-2005
4 Beck, H. E., A. I. J. M. van Dijk, A. de Roo, D. G. Miralles, T. R. McVicar,J. Schellekens, and L. A. Bruijnzeel (2016). Global-scale regionalization of hydrologic model parameters, WaterResour. Res., 52, 3599–3622, https://agupubs.onlinelibrary.wiley.com/doi/10.1002/2015WR018247