evapotranspiration; satellite; meteosat; modelling; near-real time
Abstract :
[en] The objective of VR is threefold: at first to assess the performances of proposed algorithm,
secondly to evaluate the extent to which algorithm performances conforms with requirements
targeted in Product Requirement Document (PRD) and finally, to identify error sources to open
the way towards further improvements in next versions. This report summarizes the findings
obtained during the validation of LSA-16 instantaneous MSG EvapoTranspiration (MET) and
LSA-17 Daily-accumulated MSG EvapoTranspiration (DMET) products. Three validation
approaches were adopted:
a) In-situ validation. This validation is performed by comparing the algorithm output
(instantaneous and cumulated evapotranspiration) to evapotranspiration derived from
measurements made at selected locations in different climatic/vegetation conditions.
b) Models inter-comparison. In this approach, the algorithm output is compared to the
output of models recognized to produce valuable meteorological information. In the
current validation report, data from ECMWF model and GLDAS was used to perform
models inter-comparison.
c) Consistencies check The objective is to quantify the model performances based on the
knowledge of intrinsic sources of error and to check the consistency with the other LSA-
SAF products some of which are used as input to the MET algorithm.
Based on the results of the validation tests reported in this document and its annexes, it is
concluded that overall algorithm performances are high and that LSA-SAF MET/DMET
algorithms are able to reproduce the temporal evolution of evapotranspiration (ET), with values
equivalent to observations. For estimates flagged ‘Nominal’ or ‘Below Nominal’, the PRD
quality criterion is satisfied to a rate higher than 70%. Good agreement is found for stations at
which close correspondence exist between ECOCLIMAP and station land cover, with the best
scores for stations over grass and mixed forests.
From the inter-comparison exercise it is concluded that MET estimates are in agreement with
ECMWF and GLDAS ET estimates with a spatial correlation ranging between 85% and 95%
for midday images through the studied period. For high co-zenithal angles better correlation is
found with ECMWF while for low angles (spring/late autumn and morning/evening) with
GLDAS. Observed discrepancies between models estimates are not systematic and can be
explained in terms of differences in input variables and model parameterisation. The comparison
of Tskin/LST morning heating rates highlighted regions where improvements related to soil
moisture and/or vegetation parameterisation are still possible. Concerning the validation of the
Daily ET (DMET) product, a first attempt to assess the product accuracy confirmed the results
obtained during the validation of instantaneous product; it is that spatial correlation between
images remains high (between 85 and 95%), with the highest scores over Europe.
It has been identified that the main sources of differences for the in-situ validation as well as for
models inter-comparisons are related to differences in input variables, land cover definition and
models parameterisation. Any further improvements of the MET algorithm will have positive
impact on both MET and DMET products.
