Issues with Mapping Monthly Winter
Precipitation in Near Real-Time
PRISM (Parameter-Regression Independent Slopes Model) (Daly et al. 1994, 2002, 2003) is used to spatially distribute ground based observations. For this application, it functions in climatologically-aided interpolation mode (CAI). This method takes advantage of the fact that a large portion of the spatial climate and weather patterns we see today occur repeatedly, and thus are contained in the long-term mean climatology. In CAI mode, the predictor grid is not a digital elevation model (DEM), but the official USDA 1971-2000 mean monthly climatology developed for the USDA Natural Resource Conservation Service. CAI mode is robust to changes in data density and placement over time.
Mapping of near real-time precipitation has proven to be a challenge in the winter months over some parts of the United States, mostly due to a lack of quality ground-based data. When temperatures approach freezing, many sensors are not able to capture precipitation accurately in its frozen state. Conditions in Minnesota for January of 2004 illustrate such a case. A large snowstorm occurred Jan 25-27, covering the entire state (Minnesota Climate Monitor 2004). Observational networks failed miserably in accurately recording that event in the near-real time, with the majority of ground based observations recording monthly totals at 0-50% of actual amounts. This was not an isolated incident. Underreporting of precipitation appears to be widespread, and is the most severe where NWS/FAA ASOS (Automated Surface Observing System) is the primary obersvational network. The PRISM spatial QC system was largely ineffective, because nearly all of the observations were ASOS, and virtually no reasonable data were availalble nearby.
This report documents the difficulties that are being encountered in near real-time precipitation mapping, in the hope that such information will stimulate action to improve precipitation observations and increase the availability of high-quality data for mapping efforts.
Except for the western US, most COOP stations are unavailable to SCAS for 30 days after the end of the month (e.g. January data are available around March 1st). A first run of non-quality controlled digitized data is made available by the Western Regional Climate Center for the western states within 5 days after the end of the month.
SNOTEL data are provided online by the Natural Resources Conservation Service. This data source alone provides the best network of upper elevation precipitation observations in the inter-mountain West.
The ASOS network is found at all major and minor airports, and is national in scope.
Some of the MESONETs provide high-quality observations. One such example is the Oklahoma MESONET, which alone captures a better representation of precipitation that all other data sources in the state combined.
The remaining observational networks help provide better spatial coverage. Station coverage maps for SCAS near real-time precipitation mapping can be viewed at http://www.ocs.orst.edu/prism/products/matrix.phtml
Part of the issue in near real-time modelling is simply due to the availability of data in a timely manner. The following table illustrates the time lag SCAS experiences in receiving data from each network after the last day of each month.
Data Strengths and Weaknesses
SNOTEL is the single most important network for observations at higher elevations in the snow zone. Unfortunately, since the network was designed for water supply forecasting, most of the sites are only found in the western US. This does not help when modelling the central and eastern US.
While ASOS has good spatial coverage, precipitation observations are often suspicious. Precipitation is available in 1, 3, 6, and 24 hour accumulations. Often the accumulations don't match each other. For example, summing six 1-hourly observations does not match the 6-hourly summary provided in the data report. In addition, heated tipping buckets are used to measure precipitation, and that type of gauge is known to perform poorly when measuring frozen precipitation (this appears to be the main problem behind mssing the Minnesota snow storm of January 2004).
While we aren't as familiar with the other networks, they add spatial density to the observations. At this point in time, use of the other data sources seems to improve the quality of maps produced.
Case Study Maps
Daly, C., E.H. Helmer, and M. Quinones. 2003. Mapping the climate of
Puerto Rico, Vieques, and Culebra. International Journal of Climatology,
Daly, C., R.P. Neilson, and D.L. Phillips. 1994. A statistical-topographic model for mapping climatological precipitation over mountainous terrain. Journal of Applied Meteorology 33: 140-158.
Minnesota Climate Monitor. 2004. URL:<http://climate.umn.edu/doc/journal/hc0402.htm>
Copyright (c) 2004, Spatial Climate Analysis Service, Oregon State University