Item

Abstract

The land surface forms an essential part of the climate system. It interacts with the atmosphere through the exchange
of water and energy and hence influences weather and
climate, as well as their predictability. Correspondingly, the
land surface model (LSM) is an essential part of any weather
forecasting system. LSMs rely on partly poorly constrained
parameters, due to sparse land surface observations.With the
use of newly available land surface temperature observations,
we show in this study that novel satellite-derived datasets
help improve LSM configuration, and hence can contribute
to improved weather predictability.
We use the Hydrology Tiled ECMWF Scheme of Surface
Exchanges over Land (HTESSEL) and validate it comprehensively
against an array of Earth observation reference
datasets, including the new land surface temperature product.
This reveals satisfactory model performance in terms of hydrology
but poor performance in terms of land surface temperature.
This is due to inconsistencies of process representations
in the model as identified from an analysis of perturbed
parameter simulations.We show that HTESSEL can be more
robustly calibrated with multiple instead of single reference
datasets as this mitigates the impact of the structural inconsistencies.
Finally, performing coupled global weather forecasts,
we find that a more robust calibration of HTESSEL
also contributes to improved weather forecast skills.
In summary, new satellite-based Earth observations are
shown to enhance the multi-dataset calibration of LSMs,
thereby improving the representation of insufficiently captured
processes, advancing weather predictability, and understanding
of climate system feedbacks.