Assessment of a Delay-Doppler Map (DDM) Observation Operator for use in CYGNSS Data Assimilation

Ocean wind measurements from CYGNSS have the potential to improve the observation and analysis of tropical cyclones globally. The standard Level-2 wind product, however, is defined by the 25-km spatial resolution requirement using only 15 out of a total of 187 available delay-Doppler bins. CYGNSS Level 1a data consists of a calibrated delay-Doppler map (DDM), representing the scattered power from a large area (typically hundreds of kilometers) of the ocean surface, convolved with the ambiguity function of the transmitted GNSS signal. A forward model relating the DDM to surface winds is also strongly dependent on the response and many confounding variables, such as antenna gains and GNSS satellite attitude. With such a complex observation model, it is not surprising that wind retrievals derived from DDM observables near the specular point have shown biases.

We have studied an alternative approach, direct assimilation of the Level 1a DDM product into a 2-dimensional, Variational vector wind Analysis Method (VAM). Through working at a data product much closer to the signal actually captured by the instrument, assimilation offers the potential to incorporating a representation of the instrument response into the cost function itself, rather than attempting to remove these effects through calibration. Furthermore, this method will extract information from multiple “looks” at the same specular point, making better use of data that would otherwise have been discarded.

An operator (consisting of the forward model and Jacobian) have been derived for the CYGNSS DDM Level 1a product. Early comparisons with a CYGNSS Level 1a data, obtained around several Tropical Storms, covering conditions of moderate (~ 8m/s) and high (~40 m/s) wind speeds, and exhibiting nearly symmetric as well as highly asymmetric functional shapes, have shown good agreement. A closed-form representation of the Jacobian has been derived, and shown to agree with finite difference tests. Integration of this model into existing VAM code has begun, and we expect to present the status of those experiments at the meeting, along with an assessment of the expected quality of the resulting wind product. A draft Interface Control Document (ICD) defining the model state variables and meta data has been prepared, to further future utilization of the operator.