Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

Image Modified

For EFAS v4.0, LISFLOOD was calibrated for the first time using 6-hourly model computation steps and using a dataset of 6-hourly forcings generated from in-situ observations (precipitation, average temperature, potential evaporation rate from free water surface, bare soil surface and evapo-transpiration for reference crop surface) on the period 1990-2017 (see EFAS meteorological forcing and land surface data). A set of 14 key parameters was used for model calibration and the modified Kling-Gupta efficiency criteria (KGE'; Gupta et al., 2009Kling et al., 2012) was used to evaluate the performance of hydrological simulations against observed discharge at 1137 river gauges. An Evolutionary Algorithm (EA) and ECMWF HPC computing facilities were used to perform LISFLOOD calibration.

LISFLOOD calibration was last updated in EFAS v3.0. The main improvements in LISFLOOD calibration for EFAS v4.0 include:

  • 6-hourly model computation steps;
  • improvements to hydraulic routing;
  • improvements to the meteorological forcings (better spatio-temporal data coverage, increased temporal resolution and improved data interpolation);
  • some corrections to the model rivers network;
  • improved localisation of river gauges on the model rivers network;
  • increased number of calibration points (up to 1137 from 717);
  • increased number of calibrated catchments (up to 215) and calibrated domain area (up to  9'174'725 km2);
  • 6-hourly calibration points (406 over 1137);
  • longer time series of discharge observations;
  • more strictly physical ranges for calibration parameters.

Hydrological skill was evaluated using KGE' (the modified Kling-Gupta efficiency criteria) on the period 1991-2017 (year 1990 was not included in evaluation): 60 % of stations score a KGE' higher than 0.7 on the full period of available discharge data, with a slightly higher percentage for 6-hourly stations. The spatial pattern of KGE' is substantially homogeneous, with larger rivers scoring higher KGE' and poorer performance in catchments with regulated river regimes (i.e. specially in the Iberian Peninsula).