Download presentation
Presentation is loading. Please wait.
Published byDelilah Owen Modified over 9 years ago
1
Development of a 103-Year High- Resolution Climate Data Set for the Conterminous United States Wayne Gibson 1, Christopher Daly 1, Tim Kittel 2, Doug Nychka 2, Craig Johns 2, Nan Rosenbloom 2, Alan McNab 3, and George Taylor 1 1 Spatial Climate Analysis Service, Oregon State University, Corvallis, OR 97331, USA 2 National Center for Atmospheric Research, Boulder, CO 80307, USA 3 National Climatic Data Center, Asheville, NC 29901, USA
2
Introduction Why is the data set so useful? uniqueunique complete in space and time for long time period (US, 103 years)complete in space and time for long time period (US, 103 years) high resolution (4km)high resolution (4km) spatial QC of the station data prior to modelingspatial QC of the station data prior to modeling many applications need this type of datamany applications need this type of data Methodology used to create grids Statistical infilling of incomplete station data (NCAR)Statistical infilling of incomplete station data (NCAR) PRISM model used to spatially map the station dataPRISM model used to spatially map the station data PRISM products - Official USDA 1961-1990 normals for the US -New NCDC Climate Atlas of the US (48 parameters) -Canada, China, European Alps, Pacific Islands, Puerto Rico
3
Project Overview Main objective To create serially complete, high quality, topographically sensitive, high resolution grids for the conterminous United States (ppt, Tmin, and Tmax)To create serially complete, high quality, topographically sensitive, high resolution grids for the conterminous United States (ppt, Tmin, and Tmax) To create a serially complete infilled station data setTo create a serially complete infilled station data set Progression Progression Year 1: Preliminary precipitation grids created for 1948-1993Year 1: Preliminary precipitation grids created for 1948-1993 Year 2: Development of a semi-automated Quality Control (QC) system (ASSAY QC, based on PRISM)Year 2: Development of a semi-automated Quality Control (QC) system (ASSAY QC, based on PRISM) Year 3: Development of a more robust methodology for station data infilling (National Center for Atmospheric Research)Year 3: Development of a more robust methodology for station data infilling (National Center for Atmospheric Research) Year 4: Creation of final grids for the time period 1895-1997Year 4: Creation of final grids for the time period 1895-1997
4
Collection of Station Data HCN: Historical Climate Network (1895-1997) COOP: National Weather Service Cooperative Network (1895-1997) MCC: COOP data from the Midwestern Climate Center (1895-1947) SNOTEL: SNOwpack TELemetry Network, National Resource Conservation Service (1978-1997) AG: Agricultural climate data (1961-1993) MISC: Miscellaneous data (storage gauges, snow course) Inconsistencies between Station Data Networks
5
Observation Networks over Time
6
Observation Networks vs Elevation
7
Data QC: Station Metadata Checks Elevation Using Geographical Information Systems (GIS)Using Geographical Information Systems (GIS) Latitude, longitude, and elevation: Analyzed each stations metadata for changesAnalyzed each stations metadata for changes Total of 100 metadata errors Horizontal position errors <= 2 degreesHorizontal position errors <= 2 degrees Elevation errors <= 1200 mElevation errors <= 1200 m
8
Data QC: PRISM based QC system – ASSAY QC What is Bad Data Data having transcription errorsData having transcription errors We are not attempting to identify errors such as gauge under catch, observation methods, or instrumentation changes.We are not attempting to identify errors such as gauge under catch, observation methods, or instrumentation changes. ASSAY QC – automated method Jackknifed predictionJackknifed prediction Compare predicted to observed valueCompare predicted to observed value Tag large differences as “candidate” outliers.Tag large differences as “candidate” outliers. Process of evaluating detection of outliers with actual station data (monthly and daily observations)Process of evaluating detection of outliers with actual station data (monthly and daily observations) Post Processing Additional Check to “Candidates” Applied Based on Closest/Highest Weighted Station:Additional Check to “Candidates” Applied Based on Closest/Highest Weighted Station: - Distance - Elevation - Precipitation Amount - Large Outliers in the Observations Manual ChecksManual Checks List of “Bad” Observation. Mark as Missing.
9
QC Results - Precipitation 2371 monthly data errors out of 6,345,675 station-months for a detection rate of 0.0374% This are about 2 errors per monthly grid, not insignificant Also keep in mind that there is a propagation of errors in space. (50km radius or greater)
10
Example Outlier
14
Issues Inconsistencies among Observation Networks SNOTEL vs COOP (ppt)SNOTEL vs COOP (ppt) HCN vs COOP (adjusted vs raw)HCN vs COOP (adjusted vs raw) Station data infilling errors Climatologically aided interpolation (climate as predictor)Climatologically aided interpolation (climate as predictor) ASSAY QC improvements Easily detects outliersEasily detects outliers Run iterativelyRun iteratively Independent evaluation Long term runoffLong term runoff
15
Fullerton: Location to Other Stations
17
HCN vs COOP: Inconsistencies
19
Summary Important data set. High quality, high resolution, and long duration. Important data set. High quality, high resolution, and long duration. Can be used to support a variety of research topics in many disciplines. Can be used to support a variety of research topics in many disciplines. Precipitation, Tmin, and Tmax Precipitation, Tmin, and Tmax – Summer 2002 ftp://ftp.ncdc.noaa.gov/pub/data/prism100 ftp://ftp.ncdc.noaa.gov/pub/data/prism100 http://www.ocs.oregonstate.edu/prism
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.