The Use of Synthetic Aperture Radar-Derived Wind Speed in Numerical Weather Prediction Error Detection

Satellite-borne synthetic aperture radar-derived wind speed (SDWS) images provide a means of mapping ocean surface wind fields at resolutions fine enough to quantify many boundary layer flow structures over swaths of mesoscale width. Thus, the current and projected fleets of polar-orbiting synthetic aperture radar (SAR) satellites provide the opportunity to assess the performance of numerical weather prediction models (NWP) on fine scales that can't be achieved with either satellite-borne scatterometers or in situ observations. This capability is becoming increasingly important as research, and operational, NWP models are achieving resolutions needed to capture the full spectrum of mesoscale weather phenomena and a number of boundary layer phenomena as well. This capability was evaluated in the current study.

SDWS images were compared to Weather Research and Forecasting (WRF) output for a number of intense meteorological phenomena in order to assess WRF wind field errors and how these errors varied with both the scale of the phenomena in question and with model resolution. The WRF runs featured a four two-way nested grid centered on the meteorological feature of interest. Resolution for the four grids was 54, 18, 6, and 2 km. The coarsest-grid domain thus corresponds to a fairly low resolution operational global NWP model of the current generation while the next two finer-grid domains correspond to state-of-the art regional operational NWP models. The finest-grid domain corresponds to the next generation of operational regional models. For each case WRF was run in four configurations ranging from a single 54 km grid domain to the full four-domain setup described above. Thus, the effect of the two-way nesting on model versus SDWS agreement could also be assessed.

Fifteen cases for the Gulf of Alaska and Bering Sea were examined. Mesoscale features sampled include gap flows, barrier jets, synoptic fronts, pre-frontal jets, secluded cyclones and terrain-induced gravity waves. The results suggest that WRF wind speed and wind direction errors result from both the misplacement of synoptic-scale weather features and the failure to resolve mesoscale structures in the boundary layer wind field. Increased WRF resolution improved the positioning of these features but not their intensity. Likewise, two-way nesting had only minor impact on the forecasts produced by the coarser-resolution grids.

These results confirm the hypothesis that SDWS can be used for both quantitative and qualitative analysis of the scale- and resolution-dependence of NWP model boundary layer wind forecast errors over the ocean. Moreover, SDWS images can be used to alert forecasters to feature misplacement in the early lead times of an NWP model run and to enhance forecaster understanding of which mesoscale phenomena remain unresolved by the NWP model and the synoptic settings under which those phenomena occur.