Multi-Decade Analysis of Record for Hydrologic Model Calibration

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 5 February 2014: 4:15 PM
Room C210 (The Georgia World Congress Center )
David H. Kitzmiller, NOAA/NWS, Silver Spring, MD; and W. Wu, Y. Zhang, D. A. Miller, and Z. Zhang
Manuscript (1.1 MB)

Among the major challenges facing National Weather Service (NWS) hydrologists is the development of calibration data sets for distributed hydrologic models with high-spatial and temporal resolution. Such calibration procedures require, at a minimum, gridded records of precipitation and surface air temperature spanning multiple decades. The record should feature a grid mesh length ≤ 4 km and no more than a 1-hour time increment. When developed, this analysis of record (AOR) would be suitable for calibration of distributed and lumped hydrologic models, and for statistical conditioning of ensemble hydrologic model input. Available input to the AOR includes data from a surface observation network that has changed dramatically since 1960, a variety of remote sensor platforms (radar, satellite), and several numerical model reanalyses that have assimilated various combinations of these observational data. Moreover, gridded 30-year climatology records for the years 1981-2010 are now available as long-term constraints on the record. A major challenge is finding optimum uses for all these sources. This presentation will feature a proposed methodology, based on previously-published techniques, for developing a first-guess precipitation and temperature AOR covering the conterminous United States and adjacent areas of Mexico and Canada, for the period 1979-2011. The record will incorporate hourly reanalysis, gridded 4-km hourly radar and satellite precipitation inputs, monthly temperature and precipitation inputs, and 30-year climatology constraints. The first-guess will later be augmented with quality-controlled surface temperature and precipitation observations, and with radar precipitation estimates at 1-km resolution. Initial results in identifying and correcting long-term biases in the reanalysis record will be presented.