HR: 1330h
AN: B22A-0805    [PDF]
TI: Characterizing and Filling Data Gaps in ARM Measurements for Carbon Models
AU: McCord, R A
EM: mccordra@ornl.gov
AF: Oak Ridge National Laboratory, Environmental Science Division P.O. Box 2008, M.S. 6407, Oak Ridge, TN 37831-6407 United States
AU: * Hargrove, W W
EM: hnw@fire.esd.ornl.gov
AF: Oak Ridge National Laboratory, Environmental Science Division P.O. Box 2008, M.S. 6407, Oak Ridge, TN 37831-6407 United States
AU: Jager, H I
EM: jagerhi@ornl.gov
AF: Oak Ridge National Laboratory, Environmental Science Division P.O. Box 2008, M.S. 6407, Oak Ridge, TN 37831-6407 United States
AU: Brandt, C C
EM: fcb@ornl.gov
AF: Oak Ridge National Laboratory, Environmental Science Division P.O. Box 2008, M.S. 6407, Oak Ridge, TN 37831-6407 United States
AU: Hanan, N
EM: niall@nrel.colostate.edu
AF: Natural Resource Ecology Laboratory, NESB, B217 Colorado State University, Fort Collins, CO 80523 United States
AB: The Atmospheric Radiation Measurement (ARM) data archive includes many of the measurements needed by carbon modelers to predict carbon dynamics in terrestrial ecosystems, but data gaps limit the use of ARM data as input for simulation models. Because the DOE ARM Program records actual measurements, circumstances unavoidably arise when instrument and storage failures create gaps in the temporal stream of measurements. Most temporal gaps are short in duration and affect only one or a few related parameters. However, some rare failures, such as wide-area power outages or ice storms, occasionally affect many measurement streams at one or more ARM facilities simultaneously. We have statistically characterized the frequency of univariate temporal gap lengths in various ARM measurements, and have devised approaches for filling such data gaps in space and time. To make ARM measurements suitable as model input, we identified and eliminated outliers, removed values with known QA problems, aggregated the measurements to an appropriate temporal scale (hours), and filled gaps in the data record using univariate imputation methods across time and space. We have prepared a set of hourly aggregated, gap-filled products from ARM SIRS and SMOS data collected at the SGP site from 1996 through 2001. These products were designed to facilitate the use of ARM measurements as climate drivers for carbon simulations. In cases where no raw data were available, we imputed a replacement value from adjacent hours or sites. ARM measurements differed widely in predictability. Temperature and vapor pressure were easiest to impute, but precipitation was a challenge. Shortwave radiation was more difficult to impute than longwave radiation. Successful imputation created reasonable values and patterns that were indistinguishable from the surrounding measurements. The difficulty of imputation for each measurement could help prioritize instrument repair and operational triage during data collection.
UR: http://www.archive.arm.gov/Carbon
DE: 0315 Biosphere/atmosphere interactions
DE: 0614 Biological effects
DE: 1615 Biogeochemical processes (4805)
DE: 1655 Water cycles (1836)
DE: 1851 Plant ecology
SC: Biogeosciences [B]
MN: 2003 Fall Meeting