Bayesian Non-Stationary Index Gauge Modeling of Gridded Precipitation Extremes
Abstract
We propose a Bayesian non-stationary model to generate watershed scale gridded estimates of extreme precipitation return levels. The Climate Hazards Group Infrared Precipitation with Stations (CHIRPS) dataset is used to obtain gridded seasonal precipitation extremes over the Taylor Park watershed in Colorado for the period 1981-2016. For each year, grid cells within the Taylor Park watershed are aggregated to a representative "index gauge," which is input to the model. Precipitation-frequency curves for the index gauge are estimated for each year, using climate variables with significant teleconnections as proxies. Such proxies enable short-term forecasting of extremes for the upcoming season. Disaggregation ratios of the index gauge to the grid cells within the watershed are computed for each year and preserved to translate the index gauge precipitation-frequency curve to gridded precipitation-frequency maps for select return periods. Gridded precipitation-frequency maps are of the same spatial resolution as CHIRPS (0.05° x 0.05°). We verify that the disaggregation method preserves spatial coherency of extremes in the Taylor Park watershed. Validation of the index gauge extreme precipitation-frequency method consists of ensuring extreme value statistics are preserved on a grid cell basis. To this end, a non-stationary extreme precipitation-frequency analysis is performed on each grid cell individually, and the resulting frequency curves are compared to those produced by the index gauge disaggregation method.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2017
- Bibcode:
- 2017AGUFMGC43F1117V
- Keywords:
-
- 1616 Climate variability;
- GLOBAL CHANGE;
- 1630 Impacts of global change;
- GLOBAL CHANGE;
- 1637 Regional climate change;
- GLOBAL CHANGE;
- 4313 Extreme events;
- NATURAL HAZARDS