Wednesday, 25 January 2017
4E (Washington State Convention Center )
The influence of gridded, radar-based precipitation estimates in hydrologic applications is ever increasing, but these datasets routinely suffer from biases that lead to inaccurate estimates. There are several types of biases, such as beam blockage, mean-field biases, and range-dependent biases, which have been well-studied with reliable solutions developed to improve precipitation estimates in affected regions. However, these methods typically require radar-specific information to identify the spatial structures of the biases and corrections to the precipitation fields are typically done at daily and sub-daily time scales. These correction algorithms can overlook two-dimensional spatial anisotropies in bias fields that appear on longer time scales. Additionally, corrections are applied at the individual radar or regional level, which may lead to discontinuities in adjusted precipitation fields.
We developed a data assimilation technique to improve identification of spatial anisotropies and discontinuities in gridded, radar-based precipitation estimates. Our initial precipitation fields are those adjusted using methods that have identified and adjusted for more well-behaved bias structures. This initial field is adjusted by a two dimensional bias field that is found by Ordinary Kriging interpolation of gauge-based biases. Holdout cross-validation has shown that the data assimilation technique improves the estimates by about 5% in our study region.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner