Development of high-resolution L4 ocean wind products

Abstract

[eng] Heat, moisture, gas, and momentum exchanges at the oceanic and atmospheric interface modulate, inter alia, the Earth’s heat and carbon budgets, global circulation, and dynamical modes. Sea surface winds are fundamental to these exchanges and, as such, play a major role in the evolution and dynamics of the Earth’s climate. For ocean and atmospheric modeling purposes, and for their coupling, accurate sea-surface winds are therefore crucial to properly estimate these turbulent fluxes. Over the last decades, as numerical models became more sophisticated, the requirements for higher temporal and spatial resolution ocean forcing products grew. Sea surface winds from numerical weather prediction (NWP) models provide a convenient temporal and spatial coverage to force ocean models, and for that they are extensively used, e.g., the European Centre for Medium-range Weather Forecasts (ECMWF) latest reanalysis, ERA5, with ubiquitous hourly estimates of sea-surface wind available globally on a 30-km spatial grid. However, local systematic errors have been reported in global NWP fields using collocated scatterometer observations as reference. These rather persistent errors are associated with physical processes that are absent or misrepresented by the NWP models, e.g., strong current effects like the Western Boundary Current Systems (highly stationary), wind effects as- sociated with the oceanic mesoscale (sea surface temperature gradients), coastal effects (land see breezes, katabatic winds), Planetary Boundary Layer parameterization errors, and large-scale circulation effects, such as those associated with moist convection areas. In contrast, the ocean surface vector wind or wind stress derived from scatterometers, although intrinsically limited by temporal and spatial sampling, exhibits considerable spatial detail and accuracy. The latter has an effective resolution of 25 km while that of NWP models is of 150 km. Consequently, the biases between the two mostly represent the physical processes unresolved by NWP models. In this thesis, a high-resolution ocean surface wind forcing, the so-called ERAú, that combines the strengths of both the scatterometer observations and of the atmospheric model wind fields is created using a scatterometer-based local NWP wind vector model bias correction. ERAú stress equivalent wind (U10S) is generated by means of a geolocated scatterometer-based correction applied separately to two different ECMWF reanalyses, the nowadays obsolete ERA-interim (ERAi) and the most recent ERA5. Several ERAú configurations using complementary scatterometer data accumulated over different temporal windows (TW) are generated and verified against independent wind sources (scatterometer and moored buoys), through statistical and spectral analysis of spatial structures. The newly developed method successfully corrects for local wind vector biases in the reanalysis output, particularly in open ocean regions, by introducing the oceanic mesoscales captured by the scatterometers into the ERAi/ERA5 NWP reanalyses. However, the effectiveness of the method is intrinsically dependent on regional scatterometer sampling, wind variability and local bias persistence. The optimal ERAú uses multiple complementary scatterometers and a 3-day TW. Bias patterns are the same for ERAi and ERA5 SC to the reanalyses, though the latter shows smaller bias amplitudes and hence smaller error variance reduction differences in verification (up to 8% globally). However, because of ERA5

being more accurate than ERAi, ERAú derived from ERA5 turns out to be the highest quality product. ERAú ocean forcing does not enhance the sensitivity in global circulation models to highly localized transient events, however it improves large-scale ocean simulations, where large- scale corrections are relevant. Besides ocean forcing studies, the developed methodology can be further applied to improve scatterometer wind data assimilation by accounting for the persistent model biases. In addition, since the biases can be associated with misrepresented processes and parmeterizations, empirical predictors of these biases can be developed for use in forecasting and to improve the dynamical closure and parameterizations in coupled ocean-atmosphere models.

[spa] Los vientos de la superficie del mar son fundamentales para estimar los flujos de calor y momento en la interfaz oceánica-atmosfera, ocupando un papel importante en la evolución y la dinámica del clima del planeta. Por tanto, en modelación (oceánica y atmosférica), vientos de calidad son cruciales para estimar adecuadamente estos flujos turbulentos. Vientos de la superficie del mar de salidas de modelos de predicción numérica del tiempo (NWP) proporcionan una cobertura temporal y espacial conveniente para forzar los modelos oceánicos, y todavía se utilizan ampliamente. Sin embargo, se han documentado errores sistemáticos locales en campos de NWP globales utilizando observaciones de dispersómetros co-ubicados como referencia (asociados con procesos físicos que ausentes o mal representados por los modelos). Al contrario, el viento de la superficie del mar derivado de los dispersómetros, aunque intrínsecamente limitado por el muestreo temporal y espacial, presenta una precisión y un detalle espacial considerables. Consecuentemente, los sesgos entre los dos representan principalmente los procesos físicos no resueltos por los modelos NWP. En esta tesis, se crea un producto de forzamiento del viento en la superficie del océano de alta resolución, el ERAú. ERAú se genera con una corrección media basada en diferencias geolocalizadas entre dispersometro y modelo, aplicadas por separado a dos reanálisis diferentes, el ERA-interim (ERAi) y el ERA5. Varias configuraciones de ERAú utilizando datos de dispersómetros complementarios acumulados en diferentes ventanas tempo- rales (TW) se generan y validan frente a datos de viento independientes, a través de análisis estadísticos y espectrales de estructuras espaciales. El método corrige con éxito los sesgos del vector de viento local de la reanálisis. Sin embargo, su eficacia depende del muestreo del dispersómetro regional, la variabilidad del viento y la persistencia del sesgo local. El ERAú óptimo utiliza múltiples dispersómetros complementarios y un TW de 3 días. Las dos reanálisis muestran los mismos patrones de sesgo en la SC, debido a que ERA5 es más preciso que ERAi, ERAú derivado de ERA5 es el producto de mayor calidad. El forzamiento oceánico ERAú mejora las simulaciones oceánicas a gran escala, donde las correcciones a gran escala son relevantes.