H212-06
Parameter Estimation of Distributed Models Made Easy by the MPR Tool

Wednesday, 16 December 2020: 17:45
Virtual
Robert Schweppe1, Stephan Thober2, Matthias Kelbling2, Rohini Kumar3, Sabine Attinger3 and Luis Samaniego2, (1)Helmholtz Centre for Environmental Research UFZ Leipzig, Computational Hydrosystems, Leipzig, Germany, (2)Helmholtz Centre for Environmental Research - UFZ, Computational Hydrosystems, Leipzig, Germany, (3)Helmholtz Centre for Environmental Research GmbH – UFZ, Leipzig, Germany, Computational Hydrosystems, Leipzig, Germany
Abstract:
In the community of environmental modelling, the advent of hyper-resolution Earth observations and datasets in conjuncture with growing computational resources lead to an increase in model resolution.
The mathematical representations of biogeophysical processes need to be solved for billions of grid cells and thousands of time points.
Each process requires parameters that are often neither readily available nor sensibly set to a global fixed value nor easily calibrated.
Instead they often must be inferred directly from the land surface properties through transfer functions and then scaled to the spatiotemporal modelling domain.

This two-step approach is the Multiscale Parameter Regionalization (MPR), initially introduced as part of the mesoscale hydrologic model (mHM, Samaniego et al. 2010, Kumar et al.2013).
To apply MPR to state-of-the-art environmental models that often use default, static, or classified parameters, we wrote an object-oriented and flexible Fortran library (https://git.ufz.de/chs/MPR).
It allows users to flexible construct hierarchical interdependency trees linking the target estimated effective model parameters to predictor variables through transfer functions and scaling operators.
As such, it introduces a protocol for ensuring a transparent, reproducible and flexible parameter estimation.
We demonstrate its versatility by estimating parameters for models from various disciplines (e.g. mHM, Noah-MP, HTessel) which has a significant impact on long-term key environmental states and fluxes by changing their spatiotemporal patterns and also their magnitude by up to +/- 25%.

References:
Samaniego L., et al. https://doi.org/10.1029/2008WR007327
Kumar, R., et al. https://doi.org/10.1029/2012WR012195