In autumn 2017, ECMWF started an internal project focusing on ‘Understanding uncertainties in surface–atmosphere exchange’ (USURF). The aim was to investigate the systematic forecast biases in near-surface weather parameters, disentangle their sources and identify ways to reduce them in the future. Three years on, as this investigation has drawn to a close, its main conclusions and recommendations for future model development, verification and diagnostic work have been summarised in a recent ECMWF Technical Memorandum (https://bit.ly/387SXyu). They were also presented to the autumn sessions of ECMWF’s Scientific and Technical Advisory Committees and received positive feedback from representatives of our Member and Co-operating States.
The work carried out in USURF is of interest to ECMWF’s Member and Co-operating States not only because it aims to improve forecasts of near-surface weather parameters, but also because systematic biases in these parameters are one of the major open issues in the wider weather and climate modelling communities. These biases, which manifest themselves at all forecast ranges, limit predictive skill from hours to seasons ahead. Eliminating or at least reducing them is becoming increasingly important in the context of an enhanced demand for more accurate near-surface weather forecasts. This demand is driven by various interests, such as renewable energy applications or the occurrence of more intense and frequent extreme events.
Systematic biases in near-surface temperature, humidity or winds are the result of a complex interplay between (i) processes parametrized in the atmospheric and surface columns of the forecasting system, which can lead to locally generated errors, and (ii) advection, which constitutes a non-local source of errors. Understanding the leading causes of these systematic errors, which often have complicated geographical patterns and temporal structure, is a necessary step to improve the realism of the model and enhance the near-surface forecast accuracy.
Substantial progress in this direction has been made in the USURF project by using a combination of methods, including conditional verification, process-based diagnostics at observational supersites and model sensitivity experiments. It was shown that, although near-surface forecasts have gradually improved in the past decades, systematic biases with often complicated spatio-temporal patterns remain. As detailed in the Technical Memorandum, answers were found to a number of questions related to the sources of biases in temperature (e.g. cold/warm bias over southern/northern Europe during wintertime), dew point and winds. A work plan to address some of these issues was defined. This will include, among other developments, an improved representation of the snow (figure) and a revision of the land cover and vegetation maps, accompanied by a retuning of uncertain parameters in the surface–atmosphere coupling.
USURF has also provided further evidence that increases in near-surface forecast skill in a complex Earth system model critically depend on two things: the availability of comprehensive observations, and in-depth studies using process-based diagnostics that can correctly attribute model error. Ongoing improvements to the diagnostic and verification tools used at ECMWF are therefore an important contribution towards further enhancements of forecast skill, alongside model developments.
Advances in the surface–atmosphere interactions and hydrometeorological processes at increasingly high resolutions towards kilometric scales are relevant not only to ECMWF but equally to all Member and Co-operating States. To accelerate progress, work on related topics could be carried out in partnership with colleagues from the Member and Co-operation States and could for example be an area of interest for secondments at ECMWF.